When CopyTexSubImage2D calls were clipped against the bounds of the
read framebuffer by the underlying renderer backends, the robust
resource initialization code assumed that the original destination
area would be overwritten by the renderer, which was not the case.

Add new tests which were previously failing on macOS with the
ES2_OpenGL and ES3_OpenGL backends. The Metal backend is still failing
and a follow-on bug has been filed.

Bug: angleproject:4504
Change-Id: I34821dd90597f31b3cbf0921b94756556e485c91
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2116873Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Commit-Queue: Kenneth Russell <kbr@chromium.org>