Add CPU RoCE to playback emulator

[eliotheinrich]

Description

This PR adds a CPU RoCE session backend to the playback emulator.

Docs will follow (#835).

Runtime / performance impact

N/A

Self-review checklist

Please confirm each item before requesting review. Check [x] or strike
through and explain.

Before requesting review

• I reviewed my own full diff in GitHub or my editor.
• PR is in Draft if it is not yet ready for review.
• Temporary / debugging changes have been removed.
• Local test logs reviewed; no unexplained warnings or errors.
• CI logs reviewed; no unexplained warnings or errors.
• Full CI has been run.

Scope and size

• PR is under ~1000 lines, or an exception is justified in the description.
• Refactoring-only changes are isolated in their own PR(s).
• No existing tests were disabled or modified just to make this PR pass
  (if so, an issue has been raised).

Tests

• New functionality has new tests.
• Tests fail if the new functionality is broken (including crashes), not
  just when it is missing.
• Negative tests added where exceptions are expected.
• Truth data added where simple EXPECT_* / assert checks are
  insufficient for algorithmic correctness.
• CI runtime impact considered; team notified if significant.

Documentation

• Public-facing APIs have Doxygen docs.
• User-visible behavior changes have public docs, or a follow-up is
  tracked.

Code style

• Naming follows the existing convention (snake_case vs camelCase) for
  the area being modified.

Dependencies

• No new third-party dependencies, or the team has been notified and
  OSRB tickets filed.

[vedika-saravanan]

CUDA-Q’s existing RingSlotChannel serializes response-bearing calls and exits the dispatch on a timeout, while this implementation frees timed-out slots and continues publishing requests. A late unmatched response could therefore clear slot_owner_[slot] after that slot has been assigned to a newer request. Would it make sense to update ownership only after matching the response ID and confirming that slot_owner_[it->second.slot] still equals that ID? It may also be worth adding a delayed-response-after-timeout-and-ring-wrap test.

[eliotheinrich]

Yes, good catch; addressed in 85e1b1f

[vedika-saravanan]

Could we also validate f.size <= stride_ in send()? The current playback paths check max_frame_bytes upstream, but this session is the component that owns the fixed RDMA slot size, and try_publish() later copies frame_len bytes into that slot without another bound check. Keeping the invariant at this boundary would make the backend safer to reuse and straightforward to test with a slot_size + 1 frame.

[eliotheinrich]

Addressed in 85e1b1f