Add radial_mirror's SK_SCALAR_IS_FLOAT path to radial_repeat too.

Clear out SK_SCALAR_IS_FLOAT from SkRadialGradient, and merge the logic for radial_mirror and radial_repeat.

New radial/repeated bench is 10-20x faster now using float instead of fixed; the rest are unaffected.

BUG=

Committed: http://code.google.com/p/skia/source/detail?r=11824

[mtklein, 2013-10-16 12:32:26]

There doesn't seem to be a bench for a radial gradient with a repeat tiling
mode.  Unimportant or oversight?  I only thought to do this after seeing its
SkFixedSqrt taking up a significant chunk of overall time in DM.

[reed1, 2013-10-16 12:38:03]

There is no more scalar-is-fixed, so that code can be removed

[mtklein, 2013-10-16 15:12:35]

On 2013/10/16 12:38:03, reed1 wrote:
> There is no more scalar-is-fixed, so that code can be removed

Have another look?

[reed1, 2013-10-16 15:24:04]

is there a change in our gms?

lgtm

https://codereview.chromium.org/26410003/diff/17001/src/effects/gradients/SkR...
File src/effects/gradients/SkRadialGradient.cpp (right):

https://codereview.chromium.org/26410003/diff/17001/src/effects/gradients/SkR...
src/effects/gradients/SkRadialGradient.cpp:49: // GCC doesn't like using static
functions as template arguments.  So force these to be non-static.
that is some kind of crazy

[commit-bot: I haz the power, 2013-10-16 18:14:18]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/mtklein@google.com/26410003/17001

[commit-bot: I haz the power, 2013-10-16 18:59:46]

Change committed as 11824