[draft][UR] Use urEnqueueNativeCommandExp for enqueue_custom_operation

[hdelan]

Use the experimental UR entrypoint urEnqueueNativeCommandExp to implement AdaptiveCPP's AdaptiveCpp_enqueue_custom_operation.

https://github.com/AdaptiveCpp/AdaptiveCpp/blob/develop/doc/enqueue-custom-operation.md

[hdelan]

Ping @illuhad . Should this be AdaptiveCpp_enqueue_custom_operation or ACPP_enqueue_custom_operation?

[illuhad]

@hdelan Great to see! We've recently renamed extensions to AdaptiveCpp_*, so it is now AdaptiveCpp_enqueue_custom_operation. https://github.com/AdaptiveCpp/AdaptiveCpp/pull/1477

[hdelan]

Brief perf results on AMD MI210 courtesy of @hjabird :

For GROMACS on MI210:

Version	ADH dodec	BenchMEM
Control	84%	96%
ACpp host task (this PR)	90%	100%
Original host task ext ({add/get}_native_events	90%	99%

% of reference performance.

[hdelan]

Ping @MartinWehking @aelovikov-intel @intel/llvm-reviewers-runtime @intel/llvm-reviewers-cuda @intel/sycl-graphs-reviewers

I need to get this merged before the end of week as I'll be on holiday next week before the GitHub cutoff.

[hdelan]

Ping @intel/llvm-reviewers-runtime would be great to get a review on this ASAP as we would like to merge today if spec issues all get resolved

[hdelan]

I have been asked to merge this before the PI removal is merged. So I think this is good to go @intel/llvm-gatekeepers

[sommerlukas]

@hdelan: I think this PR is still missing approval:

Waiting on code owner review from intel/llvm-reviewers-runtime

[hdelan]

Aha thanks @sommerlukas I missed that. Where does it say that?

[hdelan]

In that case ping @intel/llvm-reviewers-runtime

[hdelan]

Thanks @aelovikov-intel . Ping @intel/llvm-gatekeepers this can be merged