[llvm] [AMDGPU] Implement workaround for GFX11.5 export priority (PR #99273)

[Carl Ritson via llvm-commits]

================
@@ -1597,20 +1603,23 @@ def FeatureISAVersion11_5_0 : FeatureSet<
   !listconcat(FeatureISAVersion11_Common.Features,
     [FeatureSALUFloatInsts,
      FeatureDPPSrc1SGPR,
-     FeatureVGPRSingleUseHintInsts])>;
+     FeatureVGPRSingleUseHintInsts,
----------------
perlfu wrote:

> > GFX 9 generic target doesn't follow this workaround absorbing convention
> 
> What do you mean? We should have all workarounds included in gfx9-generic, if not, that's an issue. Can you please tell me which workaround you think is missing? I'll follow up on it.

Thanks!
On a further read I was tripped up by the workarounds being in `FeatureISAVersion9_0_Consumer_Common` which is included in `FeatureISAVersion9_Generic`, rather than directly being in `FeatureISAVersion9_Generic`.
Your explanation is helpful to understand that "gfx9-generic" basic means "gfx9-consumer-generic"?
i.e. `FeatureISAVersion9_0_Consumer_Common = FeatureISAVersion9_Generic = "gfx9-generic"`.
With the MI parts not being in this target, but still apply some of the same workarounds.

I had also forgotten binaries are built for multiple targets in the compute chain; hence why you don't want more targets.

https://github.com/llvm/llvm-project/pull/99273