Fix a bug in Save-Restore no-op optimization.

We optimize
    Save
        SaveLayer
        Restore
    Restore
into
    NoOp
        NoOp
        NoOp
    Restore

I'm considering skipping the call to SkRecordOptimize again just to eliminate
this extra variable from landing SkRecord.  Thoughts?

BUG=skia:
R=robertphillips@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/560163002
3 files changed