[None][fix] Require Python transceiver for mixed Mamba

[VALLIS-NERIA]

@coderabbitai summary

Description

MixedMambaHybridCacheManager owns attention KV cache and Python-managed Mamba state in separate pools. Its supported disaggregated-transfer route is therefore the Python transceiver with the NIXL backend. Sending it through the C++/default runtime reaches an unsupported unified-pool path and fails late during construction.

This PR:

• rejects MixedMambaHybridCacheManager whenever the effective transceiver runtime is not PYTHON, including explicit CPP and unresolved auto/default routing;
• restores the hybrid transfer tests to the supported Mixed + Python/NIXL route;
• preserves all eight KV/conv/SSM dtype combinations;
• exercises asynchronous transfer cancellation on the supported Python route;
• adds explicit coverage for the runtime-auto fallback guard.

This PR is independent of the V2 Mamba cache-manager stack.

Test Coverage

• Mixed-manager runtime guard, eight Mixed + Python/NIXL hybrid transfer dtype combinations, and cancellation coverage (11 passed)
• pre-commit run --from-ref origin/main --to-ref HEAD

PR Checklist

• Please check this after reviewing the above items as appropriate for this PR.

[VALLIS-NERIA]

/bot run

[tensorrt-cicd]

PR_Github #59454 [ run ] triggered by Bot. Commit: f1df596 Link to invocation

[tensorrt-cicd]

PR_Github #59454 [ run ] completed with state SUCCESS. Commit: f1df596
/LLM/main/L0_MergeRequest_PR pipeline #47922 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[VALLIS-NERIA]

/bot run

[tensorrt-cicd]

PR_Github #59465 [ run ] triggered by Bot. Commit: c9f2b1f Link to invocation

[tensorrt-cicd]

PR_Github #59465 [ run ] completed with state SUCCESS. Commit: c9f2b1f
/LLM/main/L0_MergeRequest_PR pipeline #47932 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[VALLIS-NERIA]

/bot run

[tensorrt-cicd]

PR_Github #59611 [ run ] triggered by Bot. Commit: 562e6ec Link to invocation

[tensorrt-cicd]

PR_Github #59611 [ run ] completed with state SUCCESS. Commit: 562e6ec
/LLM/main/L0_MergeRequest_PR pipeline #48059 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

• Please check the failed tests and fix your PR
• If you cannot view the failures, ask the CI triggerer to share details
• Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

[VALLIS-NERIA]

This change has been folded into #16427 together with its Mixed + Python/NIXL runtime guard and test coverage. Closing this draft as superseded so review and CI can continue on the combined PR.