Add a "deep parse" mode to Av1QpParser returning the true average QP.

The normal AV1 QP parser just looks at the frame global QP specified in
the frame header OBU. However, some implementation set a par macroblock
offset relative to that global QP - so average may differ significantly.

This CL introduces a more expensive mode where the full bitstream is
parsed (but still no image data decoded), allowing the a true average
QP to be calculated.

Note that this is more computationally expensive and should only be
used for testing and validation purposes.

Bug: webrtc:496266459
Change-Id: Ie62fb4d914f893ade412cf71b124d6c1f28760d0
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/474420
Commit-Queue: Erik Språng <sprang@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#47797}