Move bound and isFinite into pathref

This CL moves members from SkPath into SkPathRef to make copying cheaper. This is mainly in support of: 

Stateful PathRenderer implementation https://codereview.chromium.org/23926019/

[robertphillips, 2013-09-23 17:02:42]

I think one of Brian or Mike needs to look at this.

Derek - I think the new code in SkPathRef::CreateTransformedCopy replicates what
was happening with the Android GenID in SkPath::transform but the old behavior
didn't make sense to me to start with. In particular, it seems like the only
time we would want to _not_ bump the GenID is when the matrix is the identity.

[reed1, 2013-09-23 17:41:30]

The old code took pains to make both of these inlineable and trivial in the fast
case (the dirtybit was clear). Now there will be a real function call everytime.

Do we feel ok that we have benches for these, and that the new code's speed is
acceptable?

[bsalomon, 2013-09-23 18:39:39]

https://codereview.chromium.org/24350006/diff/10001/include/core/SkPath.h
File include/core/SkPath.h (right):

https://codereview.chromium.org/24350006/diff/10001/include/core/SkPath.h#new...
include/core/SkPath.h:1000: void setBounds(const SkRect& rect);
Eek... I didn't realize we let clients set the bounds. It seems like we change
this behavior:

SkPath a;
SkPath b;
//...
b = a;
a.setBounds(/*...*/);
b.getBounds();

[robertphillips, 2013-09-23 18:56:42]

PTAL at patches 7/8. They move SkPathRef.h to include\core to enable inlining.
The main ugliness involved (other than making it public) is having to use and
int rather then SkPath::Verb in the growForVerb methods.

https://codereview.chromium.org/24350006/diff/10001/include/core/SkPath.h
File include/core/SkPath.h (right):

https://codereview.chromium.org/24350006/diff/10001/include/core/SkPath.h#new...
include/core/SkPath.h:1000: void setBounds(const SkRect& rect);
This is only used in SkAutoPathBoundsUpdate to avoid brute force recomputation
of the bounds.

[bsalomon, 2013-09-24 17:22:15]

On 2013/09/23 18:56:42, robertphillips wrote:
> PTAL at patches 7/8. They move SkPathRef.h to include\core to enable inlining.
> The main ugliness involved (other than making it public) is having to use and
> int rather then SkPath::Verb in the growForVerb methods.
> 
> https://codereview.chromium.org/24350006/diff/10001/include/core/SkPath.h
> File include/core/SkPath.h (right):
> 
>
https://codereview.chromium.org/24350006/diff/10001/include/core/SkPath.h#new...
> include/core/SkPath.h:1000: void setBounds(const SkRect& rect);
> This is only used in SkAutoPathBoundsUpdate to avoid brute force recomputation
> of the bounds.

lgtm

[reed1, 2013-09-24 17:32:17]

lgtm

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPath.h
File include/core/SkPath.h (right):

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPath.h#new...
include/core/SkPath.h:15: #include "SkPathRef.h"
// TODO: refactor this header to move more the impl into a .cpp

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPathRef.h
File include/core/SkPathRef.h (right):

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPathRef.h#...
include/core/SkPathRef.h:17: #include <stddef.h> // ptrdiff_t
is it time to put this sucker in SkTypes.h?

[robertphillips, 2013-09-26 12:18:44]

committed as r11467

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPath.h
File include/core/SkPath.h (right):

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPath.h#new...
include/core/SkPath.h:15: #include "SkPathRef.h"
On 2013/09/24 17:32:17, reed1 wrote:
> // TODO: refactor this header to move more the impl into a .cpp

Done - although I put the TODO in SkPathRef.h not SkPath.h

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPathRef.h
File include/core/SkPathRef.h (right):

https://codereview.chromium.org/24350006/diff/18001/include/core/SkPathRef.h#...
include/core/SkPathRef.h:17: #include <stddef.h> // ptrdiff_t
I don't think so - it isn't all that widely used in Skia yet.