Start on GM to verify removal of 32767 limit in SkBlurMaskFilter::filterRectsToNine is okay

In investigating issue 1805 (Skia Debugger crashes when profiling SkPicture from chrome - https://code.google.com/p/skia/issues/detail?id=1805) it was revealed that Skia bails out of the blur rect fast path if the rect to be blurred is larger than 32767. It seems that the nine patch path should be able to handle that case. This CL attempts to prove that.

[robertphillips, 2013-11-06 21:03:07]

Not ready for review yet. It isn't testing the cases when x or y are outside the
32767 range.

[robertphillips, 2013-11-07 15:39:48]

PTAL

https://codereview.chromium.org/60513013/diff/40001/src/effects/SkBlurMaskFil...
File src/effects/SkBlurMaskFilter.cpp (right):

https://codereview.chromium.org/60513013/diff/40001/src/effects/SkBlurMaskFil...
src/effects/SkBlurMaskFilter.cpp:217: if (SkBlurMaskFilter::kInner_BlurStyle ==
fBlurStyle ||
Turns out the fast paths don't support outer blur style either.

[Tom Hudson, 2013-11-07 15:50:06]

Hard to validate all that math in the expressions for origins[].
(Do we really want -3*kSigma instead of +3*kSigma for the right/bottom corners?)

Why the value chosen for kOffset?

[robertphillips, 2013-11-07 20:25:59]

PTAL.

[reed1, 2013-11-07 20:40:11]

lgtm

[robertphillips, 2013-11-07 22:25:43]

committed as r12181