The fence wait time was increased to 120 seconds, so we don't expect
VK_TIMEOUT to be returned by a healty GPU. This change makes it so that
when VK_TIMEOUT is returned, we will treat this as an error and
propagate it to the GLES application as a GL_INVALID_OPERATION.

It is expected that the GL application will re-initialize ANGLE which
will also reinitialize the Vulkan device (and hopefully clean up the
error). This is not our final design for ANGLE's VK_TIMEOUT handling,
since we expect to expand our Vulkan device lost error handling when
implementing the GLES 3.2 robustness requirements. This will likely
improve ANGLE's VK_TIMEOUT handling as well.

Bug: angleproject:4043
Test: Manually force VK_TIMEOUT (lower kMaxFenceWaitTimeNs)
Change-Id: I2449ad214ada176014a1702a9b3417d4498d070f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1910722
Commit-Queue: Tim Van Patten <timvp@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Cody Northrop <cnorthrop@google.com>