SK_ONCE for SkMatrix::I()

SK_ONCE for SkMatrix::I()

Going to start doing these in progressively larger and larger bulk,
but I figured the first few changes probably merit caution.

BUG=

Committed: http://code.google.com/p/skia/source/detail?r=11721

[bungeman-skia, 2013-10-10 18:44:53]

lgtm, even though it seems like there should be some means of initializing this
statically (in c++11 this is trivial). Maybe add a comment about how a constexpr
constructor would make this go away?

[mtklein, 2013-10-10 18:47:37]

On 2013/10/10 18:44:53, bungeman1 wrote:
> lgtm, even though it seems like there should be some means of initializing
this
> statically (in c++11 this is trivial). Maybe add a comment about how a
constexpr
> constructor would make this go away?

Done.

[bsalomon, 2013-10-10 20:46:10]

On 2013/10/10 18:47:37, mtklein wrote:
> On 2013/10/10 18:44:53, bungeman1 wrote:
> > lgtm, even though it seems like there should be some means of initializing
> this
> > statically (in c++11 this is trivial). Maybe add a comment about how a
> constexpr
> > constructor would make this go away?
> 
> Done.

When I first wrote this I had some terrible hack where I had a simple struct
with the same layout as SkMatrix, statically init'ed it with an init list, and
then cast it to SkMatrix. But someone or some tool whined about it :)

[commit-bot: I haz the power, 2013-10-10 21:33:15]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/mtklein@google.com/26905002/4001

[commit-bot: I haz the power, 2013-10-10 21:58:34]

Change committed as 11721