This commit adds an action named `Set Copy From' to the context menu
of the Layers Docker, if one or more clone layers are selected. The
action pops up a dialog that can change the `Copy From' property of
selected clone layers.

The dialog previews the changes to these layers.
If there is any change to the document when the dialog is open, the
changes in the dialog will be applied and the dialog closed.

We shouldn't try to end the same stroke multiple times. The patch
also makes sure KisProcessingApplicator is owned by a scoped pointer
instead of manual handling.

BUG:407520

CCBUG:393112

If the template is in a read-only folder (like an installation
location) the document would be set to ReadOnly, which is wrong.

BUG:393112

Every frame add/move/delete command issues canvas updates. We shouldn't
let these updates to run until all manipulations with m_frames are
finished. Otherwise there will be non-thread-safe access to frames
storage.

BUG:405723

KArchive's KZip cannot read all files that Quazip saves. This is
a bug in KArchive, but it breaks older versions of krita that
cannot load some files saved in 4.2 anymore. So, only use quazip
if 64 bits saving is enabled.

CCBUG:407537

Original problem: widgets do not get synthesized mouse-down and
mouse-press events until the stylus is released

Reason: if the app accepts the event, WndProc should report
that to the system (by returning true). This is the only way to
prevent Windows from starting some system-wide gestures, like
click+hold -> right button click. If we ignore the event, then
OS postpones all synthesized mouse events until the entire gesture
is completed.

The patch implements a "hackish" workaround for the original problem
by using the following rules:

1) All tablet-move events are ignored (without synthesized mouse events
   OS doesn't generate any Enter/Leave events)

2) All not-accepted tablet press- and release-events and also reported as
   ignored (without it D&D doesn't work).

3) All accepted tablet press- and release-events are reported as "accepted",
   **but** we artificially synthesize mouse events for them.

TODO: there are still two problems:

1) Perhaps this synthesizeMouseEvent() is not needed at all. But we should
   first check if Qt relies on these synthesized messages anywhere in the
   code or not.

2) If we still keep synthesizeMouseEvent(), then it should fetch actual
   tablet buttons from QGuiApplicationPrivate::tabletDevicePoint(). Right
   now it always synthesizes left-click, whatever the button was pressed/
   released.

CCBUG:406668

In previous versions of Qt (wintab impeplementation) the events were
marked by Qt::MouseEventSynthesizedBySystem flag only when they were
synthesized from touch, not from tablet events. This is what
QWindowsTabletSupport does and what QQuickWindow expects (it
filters out all synthesized events). This patch recovers the old behavior
for the new QWindowsPointerHandler tablet API implementation.

CCBUG:406668

formatDifference removes global styles from local set, this is ok for
most properties, but some defaults make it impossible to set any
style combination if we remove the property.

Undo should not be able to go before the text existed

Removes code disabling optimizations in KoGamutMask and Artistic Color
Selector. It is not needed, probably left-overs from a debugging session
I was not careful enough to clean up

We should handle CR/LF problems in text files carefully

BUG:406492

BUG:400243

This allows to change the active selected text while the editor is open
allowing for faster edit of multiple texts.

BUG:400241

When a view is closed, we should ensure status bar doesn't have
a link to it. That is okay, but we shouldn't do it manually.
Keeping status bar's links is done by KisViewManager, so let just
him do his work.

BUG:407488

We shouldn't recursively start model invalidation right inside
beginRemoveRows() call.

BUG:378656

BUG:407478

In case they are created with a transparency channel. Qt converts
transparency to black when converting to grayscale.

BUG:407478

In case of conflict in i18n, keep the version of the branch "ours"
To resolve a particular conflict, "git checkout --ours path/to/file.desktop"

clang does not use attr optimze(0). This change ensures clang gets
the correct attribute (optnone). And solves the warning generated.

This reverts commit d4715e98.

Revert until I figure out a better solution

BUG:407384

BUG:407253

Why is it so hard to handle the window title correcly? Probably
because QMainWindow/QWidget are trying to be smart and helpful.

They need special handling, since the setValue isn't overriding
the base class' setValue. They also don't need to be in libs/
widgets, since they only need one class from widgetutils.