glMapBufferRange call requires CPU and GPU synchronization. Different
drivers implement this differently. Some may choose to keep multiple
copies of data to reduce the synchronization overhead while others may
choose to wait for GPU access to finish to save memory. While this
behavior is important, it should be looked under the scope of overall
game performance and we can write a standalone test for this. When
--minimize-gpu-work is specified, we mainly care about CPU overhead of
driver logic of state tracking. For this purpose, we should get this out
of picture in order to expose the true picture of what we intended to
see. This CL always adds GL_MAP_UNSYNCHRONIZED_BIT to the access bit to
avoid driver to do synchronization or make COW.

Bug: b/184766477
Change-Id: I36228a4ed9913e26aa9ad4e8446fb42ee0182c18
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2847101Reviewed-by: Cody Northrop <cnorthrop@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Charlie Lao <cclao@google.com>