This merges branch pigment-lcms2_rgbaf32-rempt. When lcms2 supports
16 bit floating point numbers, the last two colorspaces will be
moved from ctlcs to lcms, effectively retiring ctlcs. The technology
was way cool, but it's easier to support all features in lcms.

This also changes pigment:

* if an icc-based colorspace factory is encountered for which there
are no profiles, don't assert but warn and disregard the factory
* previously, if the color conversion graph would encounter its
base alpha 8 node, we would assert because there was no factory for
alpha8.
* the colorspace traits are now made complete and split out in
separate files for easier editing.

This also adds a bunch of more or less useful variations of colorspaces
to the lcms2, as well as whole new YCbCr classes. For which there are
no free profiles available, so users will have to provide those, for
instance from the lcms2 website.

Squashed commit of the following:

commit cea6e6eb
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sun Jun 3 11:28:16 2012 +0200

    fix rgba 16 bits

commit 4dfc3e7d
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sun Jun 3 11:23:52 2012 +0200

    fix lab colorspace

commit da9d6b65
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sun Jun 3 11:13:16 2012 +0200

    fix colorspace id

commit d4f26cc8
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sun Jun 3 11:12:54 2012 +0200

    fix warning

commit 41735f8e
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sun Jun 3 11:12:40 2012 +0200

    remove KIS from headerguard

commit 1342d497
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sat Jun 2 15:32:22 2012 +0200

    cleanups

commit 33c423a9
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sat Jun 2 11:45:15 2012 +0200

    Don't delete all the colorspace stuff on app exit

    It's faster and safer.

commit b7ddefaf
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sat Jun 2 11:44:19 2012 +0200

    Color Conversion System: don't assert if there is no profile

    For some colorspaces (ycbcr) there are no free profiles available. If
    the user wants, they can add a profile and the cs will work, but until
    that time, don't crash.

commit c8240f83
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Sat Jun 2 11:31:42 2012 +0200

    Implement floating point versions of all supported color models

    This also adds ycbcr support using lcms2; however, users should bring
    their own profiles.

commit 54c10a59
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Thu May 31 17:03:27 2012 +0200

    compile fixes

commit 428291b9
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Thu May 31 16:44:16 2012 +0200

    pigment: rename the rgb traits to bgr traits

    The reason is that they actually describe a bgra pixel, not an rgba
    pixel and we need rgba traits for the floating point rgba color space.

commit a08b07b3
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Thu May 31 16:41:46 2012 +0200

    remove unused include

commit 1059be2c
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Thu May 31 16:41:00 2012 +0200

    pigment: split out the traits into separate files

    And add traits for all supported color models and typedefs for all
    (potentially) supported channel depths.

commit ec7a2770
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Thu May 31 15:42:26 2012 +0200

    remove obsolete file

commit 67e13950
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Thu May 31 12:57:53 2012 +0200

    ctlss: Remove profiles and colorspace definitions that are in lcms2

    gray32f, rgba32f, xyza32, ycbcr8, ycbcr16 can be done using lcms2.

    Also remove the unused lmsaf32 profile.

commit 57f2b022
Author: Boudewijn Rempt <boud@valdyas.org>
Date:   Thu May 31 11:44:23 2012 +0200

    lcms color engine: add an experimental rgbaf32 colorspace

    This uses the new floating point capabilities of lcms2 to create a
    hdr rgba colorspace that automatically uses all the templated functions
    of pigment.