ARMv6: Add fast path for over_n_8888_8888_ca

ARMv6: Add fast path for over_n_8888_8888_ca

Benchmark results, "before" is
* upstream/master 4b76bbfda670f9ede67d0449f3640605e1fc4df0
"after" contains the additional patches on top:
+ ARMv6: Support for very variable-hungry composite operations
+ ARMv6: Add fast path for over_n_8888_8888_ca (this patch)

lowlevel-blt-bench, over_n_8888_8888_ca, 100 iterations:

       Before          After
      Mean StdDev     Mean StdDev   Confidence   Change
L1     2.7   0.00     16.1   0.06    100.00%    +500.7%
L2     2.4   0.01     14.1   0.15    100.00%    +489.9%
M      2.3   0.00     14.3   0.01    100.00%    +510.2%
HT     2.2   0.00      9.7   0.03    100.00%    +345.0%
VT     2.2   0.00      9.4   0.02    100.00%    +333.4%
R      2.2   0.01      9.5   0.03    100.00%    +331.6%
RT     1.9   0.01      5.5   0.07    100.00%    +192.7%

At most 1 outliers rejected per test per set.

cairo-perf-trace with trimmed traces, 30 iterations:

                                    Before          After
                                   Mean StdDev     Mean StdDev   Confidence   Change
t-firefox-talos-gfx.trace          33.1   0.42     25.8   0.44    100.00%     +28.6%
t-firefox-scrolling.trace          31.4   0.11     24.8   0.12    100.00%     +26.3%
t-gnome-terminal-vim.trace         22.4   0.10     19.9   0.14    100.00%     +12.5%
t-evolution.trace                  13.9   0.07     13.0   0.05    100.00%      +6.5%
t-firefox-planet-gnome.trace       11.6   0.02     10.9   0.02    100.00%      +6.5%
t-gvim.trace                       34.0   0.21     33.2   0.21    100.00%      +2.4%
t-chromium-tabs.trace               4.9   0.02      4.9   0.02    100.00%      +1.0%
t-poppler.trace                     9.8   0.05      9.8   0.06    100.00%      +0.7%
t-firefox-canvas-swscroll.trace    32.3   0.10     32.2   0.09    100.00%      +0.4%
t-firefox-paintball.trace          18.1   0.01     18.0   0.01    100.00%      +0.3%
t-poppler-reseau.trace             22.5   0.09     22.4   0.11     99.29%      +0.3%
t-firefox-canvas.trace             18.1   0.06     18.0   0.05     99.29%      +0.2%
t-xfce4-terminal-a1.trace           4.8   0.01      4.8   0.01     99.77%      +0.2%
t-firefox-fishbowl.trace           21.2   0.03     21.2   0.04    100.00%      +0.2%
t-gnome-system-monitor.trace       17.3   0.03     17.3   0.03     99.54%      +0.1%
t-firefox-asteroids.trace          11.1   0.01     11.1   0.01    100.00%      +0.1%
t-midori-zoomed.trace               8.0   0.01      8.0   0.01     99.98%      +0.1%
t-grads-heat-map.trace              4.4   0.04      4.4   0.04     34.08%      +0.1%  (insignificant)
t-firefox-talos-svg.trace          20.6   0.03     20.6   0.04     54.06%      +0.0%  (insignificant)
t-firefox-fishtank.trace           13.2   0.01     13.2   0.01     52.81%      -0.0%  (insignificant)
t-swfdec-giant-steps.trace         14.9   0.02     14.9   0.03     85.50%      -0.1%  (insignificant)
t-firefox-chalkboard.trace         36.6   0.02     36.7   0.03    100.00%      -0.2%
t-firefox-canvas-alpha.trace       20.7   0.32     20.7   0.22     55.76%      -0.3%  (insignificant)
t-swfdec-youtube.trace              7.8   0.02      7.8   0.03    100.00%      -0.5%
t-firefox-particles.trace          27.4   0.16     27.5   0.18     99.94%      -0.6%

At most 4 outliers rejected per test per set.

Cairo perf reports the running time, but the change is computed for
operations per second instead (inverse of running time).

Confidence is based on Welch's t-test. Absolute changes less than 1%
can be accounted as measurement errors, even if statistically
significant.

v4, Pekka Paalanen <pekka.paalanen@collabora.co.uk> :
	Use pixman_asm_function instead of startfunc.
	Rebased. Re-benchmarked on Raspberry Pi.
	Commit message.

v5, Ben Avison <bavison@riscosopen.org> :
	Fixed the bug exposed in blitters-test 4928372.
	15 hours of testing, compared to the 45 minutes to hit
	the bug originally.
    Pekka Paalanen <pekka.paalanen@collabora.co.uk> :
	Squash the fix, re-benchmark on Raspberry Pi.