v3dv: Z/S clearing optimizations
This series implements two changes:
- Avoids falling back to full tile buffer clears whenever we have a Z/S TLB clear to handle. This is not required because some of the hardware bugs that caused this were not properly described or didn't affect V3D 4.2.
- Enables Early Z/S clearing when possible.
With this series I observe an average 3-5% performance improvement on the Quake ports. Early Z/S clearing also kicks in for the UE Shooter demo, however in that case I don't observe any benefits which is not surprising, since we know the bottleneck in that case are the QPUs.