Roll src/third_party/skia 499abab:21b998b

Roll src/third_party/skia 499abab:21b998b

Summary of changes available at:
https://chromium.googlesource.com/skia/+log/499abab..21b998b

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
TBR=herb@google.com

Commits in this roll:
21b998b caryclark@google.com Revert of experiment with zero-length round capped line segments (patchset #4 id:60001 of https://codereview.chromium.org/1309753002/ )
dd3c165 caryclark@google.com experiment with zero-length round capped line segments
1cbdcde joshualitt@chromium.org Fix for copy surface not handling dirty context
7f9c9eb joshualitt@chromium.org wire up reduced atlas and zero textblob cache to test
d2b23e0 joshualitt@chromium.org Standardize BW to NonAA
2244c27 joshualitt@chromium.org fill rect batch refactor into separate batches
b6b513b joshualitt@chromium.org Create separate entry points for the various flavors of drawRect
ea73fa5 herb@google.com Add instructions to authenticate for fetch.
850f2d9 robertphillips@google.com Return failure if getInputResultsGPU fails in SkXfermodeImageFilter
b4fd893 caryclark@google.com add old fix for iOS SampleApp
2f0a728 halcanary@google.com GMs: change some GMs to use SkImage over SkBitmap
b30e9d3 msarett@google.com Clean up ifdefs in SkGifCodec
a71170c wangyix@google.com This code should've been part of the CL that added registerChildProcessor(); without updating fUsesLocalCoords in the parent when a child proc is registered, batching will not work properly.
c5d58f0 vbuzinov@nvidia.com Use stencil ops that don't issue writes whenever possible

Review URL: https://codereview.chromium.org/1312683002

Cr-Commit-Position: refs/heads/master@{#344995}