[None][perf] Remove spurious sync in sparse fmha forward

[brb-nv]

Description

This MR removes spurious syncs in sparse fmha forward
Before: Blocking D2H copy due to kv_segment_lens living on GPU
After: kv_segment_lens lives on CPU as passed by TRTLLM.

No changes to MSA repo.

For ISL=8192 and OSL=1 @ c=1, request latency in ms:

Before:

[Latency] P50    : 257.6939
[Latency] P90    : 261.0175
[Latency] P95    : 261.7910
[Latency] P99    : 262.7895
[Latency] MINIMUM: 256.4940
[Latency] MAXIMUM: 262.7895
[Latency] AVERAGE: 258.1691

After:

[Latency] P50    : 249.3527
[Latency] P90    : 250.5275
[Latency] P95    : 250.7692
[Latency] P99    : 252.6669
[Latency] MINIMUM: 248.4402
[Latency] MAXIMUM: 252.6669
[Latency] AVERAGE: 249.5198

Test Coverage

$ pytest tests/integration/defs/accuracy/test_llm_api_pytorch.py::TestMiniMaxM3::test_nvfp4[use_msa=True] -s -v

PR Checklist

Please review the following before submitting your PR:

• PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

• PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

• Test cases are provided for new code paths (see test instructions)

• If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.

• Any new dependencies have been scanned for license and vulnerabilities

• CODEOWNERS updated if ownership changes

• Documentation updated as needed

• Update tava architecture diagram if there is a significant design change in PR.

• The reviewers assigned automatically/manually are appropriate for the PR.

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

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.