"Fix" for hairline corner bugs

This CL fills all the holes but will introduce some (hopefully less visible) artifacts.

[robertphillips, 2013-09-12 19:54:57]

updated hairlines GM

[bsalomon, 2013-09-12 19:56:39]

On 2013/09/12 19:54:57, robertphillips wrote:
> updated hairlines GM

lgtm

[robertphillips, 2013-09-13 13:06:51]

I verified that for bench_pictures there is no performance regression with this
change.

[commit-bot: I haz the power, 2013-09-18 21:59:21]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/robertphillips@google.com/23708036/7001

[commit-bot: I haz the power, 2013-09-18 22:03:58]

Retried try job too often on Build-Mac10.7-Clang-x86-Release-Trybot for step(s)
BuildBench, BuildEverything, BuildGm, BuildSkiaLib, BuildTests, BuildTools
http://173.255.115.253:10117/buildstatus?builder=Build-Mac10.7-Clang-x86-Rele...

[robertphillips, 2013-09-18 22:15:10]

committed as r11365

[robertphillips, 2013-09-19 02:24:26]

This change required rebaselining of the following GMs:

strokes3 hairmodes fatpathfill shadertext2 strokerect  getpostextpath
pathopsinverse dashcubics path-reverse ovals rrect roundrects patheffect
hairlines shadertext3 pathinterior samplerstress complexclip_bw complexclip_aa
complexclip_aa_layer complexclip_bw_layer

Most of these were GPU-only rebaselines except for hairlines where the GM itself
was changed (thus requiring 8888 & 565 rebaselines too)