Optimize angle::Spinlock performance This optimization is a preparation for future commits. Performance was tested by defining `ContextMutexType = angle::Spinlock` to use `SharedContextMutex<angle::Spinlock>` as `mSharedContextMutex`. Test result is average time for 1000'000 `glGetError()` calls on S906B with locked CPU frequencies: SharedContextMutex<angle::Spinlock> Before the optimization: 14.289 ms After the optimization: 12.941 ms (-9.4%) Performance when not using `angle::Spinlock` (current code): SingleContextMutex: 9.115 ms (-36.2%) SharedContextMutex<std::mutex>: 29.186 ms (+204.3%) Bug: angleproject:8226 Change-Id: I447ff807cd6b72ff3fc115e73736fe63eb642785 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4632728 Commit-Queue: Igor Nazarov <i.nazarov@samsung.com> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>