GitLab

This reduces memcpy by only storing incomplete header bytes that are
only needed to store for next round decoding rather than whole data.

This make the perf faster a bit:
(Win10, local release build not fully optimized on Z840)
ToT:  147 MB/s
This: 154 MB/s (+4.5%)

Bug: 865001
Change-Id: I52de6b178034d69d7ff538d31c2ed660504e04a3
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1715921Reviewed-by: Adam Rice <ricea@chromium.org>
Reviewed-by: Yutaka Hirano <yhirano@chromium.org>
Commit-Queue: Yoichi Osato <yoichio@chromium.org>
Cr-Commit-Position: refs/heads/master@{#681743}