[CUDA][LIBCLC] Implement RC11 seq_cst for PTX6.0

[JackAKirk]

Implement seq_cst RC11/ptx6.0 memory consistency for CUDA backend.

See https://dl.acm.org/doi/pdf/10.1145/3297858.3304043 and https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#memory-consistency-model for full details. Requires sm_70 or above. With this PR there is now a complete mapping between SYCL memory consistency model capabilities and the official CUDA model, fully exploiting CUDA capabilities when possible on supported arches.

This makes the SYCL-CTS atomic_ref tests fully pass for sm_70 on the cuda backend.

Fixes #11208

Depends on #12907

[JackAKirk]

This is ready for review. I've marked this as draft so this doesn't get merged, since the UR tag is only temporary for testing: this PR requires oneapi-src/unified-runtime#1291

Some further information:

The Repaired C++11 memory consistency model (RC11 https://pure.mpg.de/rest/items/item_2543045/component/file_3332084/content ) was adopted in C++20 definition of seq_cst: see https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p0668r5.html
Nvidia explicitly state how their ptx instructions map to RC11 here (Figure 11) : https://dl.acm.org/doi/pdf/10.1145/3297858.3304043

This PR implements this mapping.
Read "4.2 A Mapping from Scoped C++ onto PTX" for an explanation of the RMW mapping from ptx 6.0 to RC11.

The "ptx 6.0" memory model described in that paper is described as the ptx memory consistency model in their official ptx documentation https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#memory-consistency-model

CUDA memory consistency patterns falling outside the scope of "ptx 6.0", apart from its "ptx 7.5" extension (https://dl.acm.org/doi/pdf/10.1145/3470496.3533045), are not properly defined anywhere.

[Alcpz]

SYCLcompat changes look good to me. Thank you for fixing those tests.

[ldrumm]

Your PR description will become the squashed commit message. Please reword it so it better reads as one:

The title should be clearer: "Implement CRC11 seq_cst for PTX6"

The wording should use imperative mood:

s/this PR implements/Implement

s/With this PR//g

There's also a merge conflict that needs to be resolved

[JackAKirk]

Bindless images failure is unrelated. This PR just needs oneapi-src/unified-runtime#1291 merged and the UR tag updated. Then it will be ready for merge. I'll unmark it as draft when that happens.

[kbenzie]

oneapi-src/unified-runtime#1291 has been merged, however this PR is dependent on #12907 merging first. Once that's happened pull in the latest sycl branch changes, resolve the conflict, then update the UR repo/tag as suggested. After that make this ready for review and UR reviewers approve it.