Move matrix operations from X server to pixman 0.13.2, handle RandR transform matrices in floating point.

 Backported from X.org, inspired by:

 commit 9ffc6719390df8fdd0a5295a7a7a0eaea792be45
 Author: Keith Packard <keithp@keithp.com>
 Date:   Mon Nov 24 13:08:48 2008 -0800

    Move matrix operations from X server to pixman 0.13.2

    pixman 0.13.2 now holds all of the matrix operations. This leaves
    the protocol conversion routines and some ABI stubs in place
Signed-off-by: Keith Packard <keithp@keithp.com>

 commit c4b9ab6bf56139fdd8c7c584a6f523c6766cddd6
 Author: Keith Packard <keithp@keithp.com>
 Date:   Tue Apr 29 08:29:42 2008 -0700

    Handle transform failure when computing shadow damage area.

    PictureTransformBounds can fail, when this happens, damage the entire screen
    so that the shadow gets repainted correctly.

 commit fa6a1df209bd74da1d545982cca437afc2198cc1
 Author: Keith Packard <keithp@keithp.com>
 Date:   Fri Mar 21 02:35:28 2008 -0700

    Avoid overflow in PictureTransformPoint. Fix PictureTransformIsIdentity.

    PictureTransformPoint computes homogeneous coordinates internally, but fails
    to handle intermediate values larger than 16.16. Use 64 bit intermediate
    values while computing the final result at 16.16 and only complain if that
    result is too large.

    PictureTransformIsIdentity was completely wrong -- it was not checking for
    identity transforms at all.

 commit 49db14e4ac26070ed86088419483888dda18b603
 Author: Keith Packard <keithp@keithp.com>
 Date:   Wed Mar 19 00:46:35 2008 -0700

    Handle RandR transform matrices in floating point.

    RandR matrix computations lose too much precision in fixed point;
    computations using the inverted matrix can be as much as 10 pixels off.
    Convert them to double precision values and pass those around. These API
    changes are fairly heavyweight; the official Render interface remains fixed
    point, so the fixed point matrix comes along for the ride everywhere.

 commit 97ab0c6eff870b52c0383b63a78cec49059b2545
 Author: Keith Packard <keithp@keithp.com>
 Date:   Tue Mar 18 15:15:40 2008 -0700

    When converting from double to fixed, round carefully.

    This reduces the matrix representation error after inverting a
    transformation matrix (although it doesn't eliminate it entirely).

    Perhaps we should extend Render to include 64-bit floating point transforms...

 commit 160252d94f04acc95f0a4e0f884ff565a5aa0744
 Author: Keith Packard <keithp@keithp.com>
 Date:   Mon Mar 17 23:03:56 2008 -0700

    Add matrix inversion function (uses doubles)

    The obvious matrix inversion function, coded using doubles to avoid fiddling
    with fixed point precision adventures.

 commit ff9d1cd843a9b0aba69a3d788b21d5f6af702590
 Author: Keith Packard <keithp@keithp.com>
 Date:   Thu Mar 13 21:30:18 2008 -0700

    Add funcs to convert between protocol and pixman matrices

 commit f547650328287545a7a4d96df8d6a6c606dd95a9
 Author: Keith Packard <keithp@keithp.com>
 Date:   Thu Mar 13 14:50:13 2008 -0700

    Export a bunch of matrix operations from render.

    The render extension uses many matrix operations internally, this change
    exposes those functions to other parts of the server, drivers and
    extensions. The change is motivated by the 'transform' additions to the
    RandR extension but will likely be useful elsewhere.

 Backport to nx-libs: Mike Gabriel <mike.gabriel@das-netzwerkteam.de>