b2c: disable amdgpu's gpu recovery
We are depending on hang_detection to detect when the GPU is hung. However, it is possible that the hang would only be seen on dEQP's context, but not in the hang_detection one. This leads to the hang going unoticed.
To fix this, we should have dEQP check if its context is still working, before running the next test. However, this is not doable yet with Vulkan, so in the mean time, disable gpu recovery to make sure that if any context hangs, all the others will hang too.