feat(ds4): add external ROCTX phase markers

[cheese-cakee]

Summary

• add default-off semantic ROCTX ranges for external ROCm tracing
• cover monolithic and heterogeneous prefill, speculative decode, and leaf layer ranges
• label exact prefill, decode, verification, and ambiguous remote work explicitly
• dynamically load ROCTX only when enabled and keep non-HIP builds as no-ops

This replaces the superseded historical in-process HIP-event profiler on current main. The old implementation recorded on a null/default stream and synchronized events, which could fence unrelated streams and distort the two-owner overlap being measured. This rewrite intentionally performs no internal GPU timing.

Behavior contract

DFLASH_DS4_ROCTX=1 enables balanced semantic ranges for correlation with an external rocprof kernel/copy trace. The marker layer:

• uses no HIP/CUDA events
• performs no stream or device synchronization
• calculates no elapsed time
• remains silent and does not load ROCTX when unset or false
• reports a one-time diagnostic when explicitly enabled but ROCTX cannot be loaded
• closes incomplete library handles when required symbols are missing
• balances successful pushes through RAII, including early-return paths
• avoids recursive layer-range double counting by marking leaf ranges only

Validation completed

• strict native C++17 standalone build and tests passed
• strict HIP-defined Linux standalone build and tests passed with dynamic-loader linkage
• git diff --check passed
• fresh independent final-diff review: no remaining actionable findings

Pending before ready-for-review / profiling claims

• full current-tree CUDA build and CTest
• full HIP build
• gfx1151 execution with a real rocprof marker/kernel/copy capture
• marker-to-kernel alignment inspection for monolithic and heterogeneous exact prefill
• alternating enabled/disabled runs to quantify instrumentation perturbation

The Lucebox machines are currently unavailable, so this is reopened as a draft. The completed standalone checks do not constitute target-hardware validation, and no performance result is claimed. Historical CI and reviews applied to the superseded event-timing implementation and should not be treated as evidence for this rewrite.

Stack

Independent current-main PR. It does not depend on #548 or the #569/#575/#577 performance stack. #575 may provide a useful external rocprof collection wrapper, but it is not a code prerequisite.

[cubic-dev-ai]

All reported issues were addressed across 10 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[cubic-dev-ai]

All reported issues were addressed across 6 files (changes from recent commits).

<sub>Tip: Review your code locally with the cubic CLI to iterate faster.

Re-trigger cubic</sub>

[cubic-dev-ai]

All reported issues were addressed across 5 files (changes from recent commits).

<sub>Tip: Review your code locally with the cubic CLI to iterate faster.

Re-trigger cubic</sub>

[cubic-dev-ai]

All reported issues were addressed across 10 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>