[llvm] [openmp] [OpenMP][AMDGPU] Adapt dynamic callstack sizes to HIP behavior (PR #74080)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> When using the `TargetParser` I was able to construct the FeatureMap corresponding to my present device(s) but that did not contain the info (and from what it contained, MaxWaveScratchSize does not really seem to 'fit' in that group of device capabilities).

The TargetParser does seem to only be exposing boolean-ish looking names, which unfortunately would imply introducing N different feature names there. I've lost track of why you need this in the first place; if you over-commit on the stack size, won't the lower level API fail for you?

https://github.com/llvm/llvm-project/pull/74080