Changed the number of axes displayed for the "snowflake" mode of multibush tool, now there are twice as many as for the symmetry mode.

This commit adds Spiral and Reverse Spiral gradients
to the gradient tool.

The regression was caused by c74ddaf2

BUG:292726

QFileInfo::isWritable() doesn't work on NTFS by default, so we should
just try to create a test fiel to check that.

BUG:411129
BUG:411081

This commit adds the proper call to update precision on resize during
dab cache fetch. It also adds signals to update the String in Gui if the
user changes Size of brush while auto precision is activated.

BUG:408052

Every time we decrement accumulatedScrollDelta we should decrement
it with DefaultDeltasPerStep, not with 1's :)

BUG:410940

Lockfree hash table iterators have a requirement of no-concurrent-inserts,
so we must ensure that the swapper and pooler do *not* insert any tiles,
why iteration is in progress. KisTileDataStore::trySwapTileData() is called
exactly from inside of the swapper iteration loop.

If we break this requirement, the insert operation may request a table
migration, and the iterator will continue iterating the old, invalid,
table.

The actual crash happened after line 'item = iter->next()' in
KisTileDataSwapper::pass(), which actually returned value '1', meaning
"Redirect". Address 0x0000000000000029 is a fetch of
KisTileData::m_mementoFlag (offset 0x28) with 'this' set to 0x01 :)

BUG:411081

Job coordinator of the table may have three states:

1) jobCoordinator == 0: the table was initialized, but no migration
   has ever been started for this table
2) jobCoordinator > 1: a table migration is in progress right now. The
   caller should join *jobCoordinator to help finishing the migration.
3) jobCoordinator == 1: the table migration has been finished. It means
   that the table doesn't have any values in it anymore.

The patch actually fixes the map to consider case 1 as a situation, when
there is no migration.

CCBUG:411081

The word ‘Authors’ means different things in difference places
in Krita. Here it means ‘Krita developers’, and a message context
is needed so that translators can supply an appropriate translation.

The status message textedit field was too narrow to show a status
message without breaking it over several lines. The relative widths of
the tablet tester painting area and the textedit field have therefore
been changed to make the latter wider, so that there’s enough horizontal
space for the status messages. The sizeHint for the tablet tester area
also had to be changed to make the dialogue wide enough.

When the dialogue was resized to be very small, so that there were
hardly any space for the tablet tester painting area, Krita would
freeze. Now, a minumum size of 150 × 150 pixels ensures that this
never happens.

The tablet tester dialogue’s ‘preferred’ size is being overridden
by KoDialog, resulting in a very small dialogue box, with a tiny
painting area for testing the tablet. Now we add a sizeHint for this
area, so that the dialogue automatically gets a reasonable size.

The width and height values for the sizeHint was chosen so that
they correspond to the size the widget would have had if KoDialog
didn’t override the preferred window size.

If there is a really heavy flow of events, the OS will start compressing
wheel events into bigger deltas. So we should do multiple zoom actions,
not one. It might fix the problem on Windows, but it will *not* fix the
problem on Linux, because on Linux XCB just drops all the MOTION events
for the mouse-wheel subdevice and sends DEVICE_CHANGED events instead.
And there is no easy way to generate QWheelEvent from DEVICE_CHANGED
event.

CCBUG:410940

If a gamut mask is active, it is also shown in the on-canvas popup
selector (default Shift+I).

KisCanvasResource provider stores activation state of current gamut
mask. Activation state is synchronized across color selectors and
their gamut masking toolbars.

Signed-off-by: Scott Petrovic <scottpetrovic@gmail.com>

Before this commit, if you chose Fuzzy Dab sensor
in Rotation option, the value of the Strength slider wouldn't
affect the resulting rotation of the brush tip.
This commit fixes that behaviour: now if you choose very low
Strength and the first curve from the template ones,
the brush tip will rotate only slightly (before it would rotate
within the full 360* range).

* Fuzzy Dab is 'additive' kind of sensor, so it's value would be put
in 'realAdditivePart' in rotationLikeValue(). 'constant' contains the
value of 'Strength' slider. To get it working together,
'realAdditivePart' had to be multiplied by the 'constant' value.

* HSV option just ignored the problem and multiplied it by the value of
the slider later. It couldn't be done for Rotation because Rotation
takes the canvas rotation into account (see rotationLikeValue).
Therefore it had to be done inside rotationLikeValue function, not
later. Therefore the multiplication in HSV option had to be disabled
to not double the effect.

BUG:376179
BACKPORT:krita/4.2

Some tablet drivers (Lenovo Yoga on Linux) have a bug [0] when the stylus
changes its own ID right in the middle of the stroke/hovering. Sometimes
they also generate spurious tabler press/release events. In Krita
such behavior may lead to recursive tool switching right in the middle
of TabletPress processing. Which could lead to a stroke started twice,
which basically leads to a freeze.

[0] - https://github.com/linuxwacom/input-wacom/issues/66#issuecomment-463349786
CCBUG:410797

1) Some shapes (e.g. Text Shape) may emit shape-changed signal right
   during painting. It would cause a deadlock.

2) add/removeShape methods should be more careful about locking
   because of their tail-recursion nature.

BUG:410909
BUG:410572

Tests for transform worker were failing because the result files
contained results from the old version of NN/Box filter.
Since I changed it a bit in 57c3194c
the old result files were no longer valid.
This commit updates the result files.

Also since the execution of the filter on one of the images
resulted in a transparent pixel in the location that the previous filter
had a non-transparent pixel, it changed the size of the resulting image
so it didn't pass the test partially because of the change in size.
Since other tests don't check the size, I removed the checks in this
test, too. (Because of the transparency in the test image, it is possible
that the resulting image will have a different size even though it is
not scaled in this dimension.)

BACKPORT:krita/4.2

Before this commit, if you have a preset like Fill_circle
that has "Use the same curve" checkbox checked and you were
clicking on other sensors, the curve would be back to the default
curve (the one that other sensors had) if you switch back and forth
between some sensors (like Pressure -> X-Tilt-> Pressure).

Also before this commit if you have the default uprising curve
and you switched Rotation -> Tilt elevation, the curve would change.

This commit fixes both behaviours (first by adding transferCurve(),
second by switching the line in kis_curve_widget).

This commit also includes two slot functions instead of one function
that were used as a slot for two different signals. It helps with
debugging signals and keeping stuff clear and ordered.

There is one problem left: if you switch Tilt elevation -> Distance
(using the 'hill' curve), the preset is set to modified even though
the curve doesn't change.

BUG:383909
BACKPORT:krita/4.2

BUG:406973

The patch basically makes KoShapeManager thread safe by adding
a simple mutex. The problem is that both,
KoCreatePathTool::Private::endPointAtPosition() and
KisRepaintShapeLayerLayerJob access the shape manager in different
threads concurrently, which obviously causes a crash.

BUG:410572
BACKPORT:krita/4.2

Anyway we always recalculate tree before any access to the shapes

CCBUG:410572
BACKPORT:krita/4.2

The sliders had two spaces (e.g. ‘Opacity:  75%’) instead of one
(‘Opacity: 75%’). The extra space seems to be added by accident.
They were at least not present in the original review board request
(https://git.reviewboard.kde.org/r/122946/), only in the actual commit
(6d6f154a).

We should pass the keys to the tool, even when it is not active
to make sure its keyPress/keyRelease methods work fine. To make
it works we need to initialize the environment correctly in
KisToolInvocationAction::processUnhandledEvent;

BUG:410532
BACKPORT:krita/4.2

- translate UI strings
- fix passing arguments to i18n

In some languages, the percentage suffix is written differently
than in English, often with a (non-breaking) space between the
number and the percentage sign (e.g. ‘75 %’).