Support multiple PDF rendering backends in the GM

gm/tests/outputs/pipe-playback-failure/output-expected/stderr

Index: gm/tests/outputs/pipe-playback-failure/output-expected/stderr
diff --git a/gm/tests/outputs/pipe-playback-failure/output-expected/stderr b/gm/tests/outputs/pipe-playback-failure/output-expected/stderr
index 573600f60e521bb447dee277440f49ac38b8866b..150a10b8cd63619f5ab8a8994b6c8c10f7f0486e 100644
--- a/gm/tests/outputs/pipe-playback-failure/output-expected/stderr
+++ b/gm/tests/outputs/pipe-playback-failure/output-expected/stderr
@@ -1 +1,2 @@
+GM: FAILED to write bitmap: gm/tests/outputs/pipe-playback-failure/output-actual/mismatchPath/comparison/selftest1-pipe.png

[epoger, 2013/08/14 19:45:59]

Do you know why this changed?  Seems surprising to me... this is the only stderr
file that changed in this CL.

[ducky, 2013/08/14 19:56:42]

I actually looked into this issue - it was one of the most confusing ones. My
best guess is that this failure was always happening, however, it only shows up
now that I've moved the error reporting into write_bitmap. Since the style for
the rest of the code (well, the test_drawing path, at least) seems to be check
conditions before writing the bitmap, I think this sounds reasonable. A future
refactor job might be able to address the core issue behind this.

 GM: ---- comparison/selftest1-pipe.png: 60000 (of 60000) differing pixels, max per-channel mismatch R=135 G=246 B=135 A=0