Added TODOs to StyleImage::create() to take PassRefPtrWillBeRawPtr

Changed StyleImage::create() methods to take PassRefPtrWillBeRawPtrs

Changed the create() methods in StylePendingImage and
StyleGeneratedImage to take PassRefPtrWillBeRawPtrs instead of pointers.
This requires some temporary const_cast<>s at callsites. Also added
TODOs to these methods to take const template arguments and remove the
const_casts once these are supported by RefPtrs.

BUG=526586

[sashab, 2015-10-07 04:33:13]

sashab@chromium.org changed reviewers:
+ alancutter@chromium.org

[alancutter (OOO until 2018), 2015-10-07 04:38:05]

https://codereview.chromium.org/1386413002/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/style/StyleGeneratedImage.h (right):

https://codereview.chromium.org/1386413002/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/style/StyleGeneratedImage.h:37: // TODO(sashab):
Make this take a PassRefPtrWillBeRawPtr<const CSSImageGeneratorValue>.
In the meantime we can make these
PassRefPtrWillBeRawPtr<CSSImageGeneratorValue>.

[sashab, 2015-10-07 04:55:20]

Is this what you meant? Not sure if this is a good idea... Introduces a lot of
const_cast<>s.

[alancutter (OOO until 2018), 2015-10-07 05:09:29]

On 2015/10/07 at 04:55:20, sashab wrote:
> Is this what you meant? Not sure if this is a good idea... Introduces a lot of
const_cast<>s.

Yep. Each const_cast should have a comment for why. For these cases it's to take
a ref.
We should avoid hiding ugliness in our code or we may forget to clean it up.

[sashab, 2015-10-07 06:27:13]

Ok, I think I see what you mean. PTAL.

[alancutter (OOO until 2018), 2015-10-07 06:35:55]

lgtm

[sashab, 2015-10-08 04:59:01]

sashab@chromium.org changed reviewers:
+ timloh@chromium.org

[sashab, 2015-10-08 04:59:02]

timloh thoughts welcome... not sure how you feel about this patch.

[sashab, 2015-11-03 00:09:02]

Ping timloh... been like 4 weeks