Honour --group-size when replaying a trace, and read traces from GCS

[wenxindongwork]

Two small usability fixes to the agentic benchmark, both hit while sweeping batch size against the MLPerf traces.

--group-size was ignored when replaying a trace

run_group took its stream count from len(specs), so --group-size did nothing whenever --trace-file was supplied. A run intending 8 streams silently replayed the trace's full 16, and the only visible sign was the stream count buried in the report — easy to miss, and it quietly invalidates a batch-size sweep.

Each group is now trimmed to --group-size at load time, so the startup summary and the run agree.

--trace-file now accepts gs://

python benchmark_agentic.py \
  --model Qwen/Qwen3.5-397B-A17B-FP8 \
  --trace-file gs://... \
  --global-prefix-len 6476 --num-groups 1 --group-size 8 --concurrency 1

[github-actions]

Description

Start with a short description of what the PR does and how this is a change from
the past.

The rest of the description includes relevant details and context, examples:

• why is this change being made,
• the problem being solved and any relevant context,
• why this is a good solution,
• some information about the specific implementation,
• shortcomings of the solution and possible future improvements.

If the change fixes a Github issue, please include a link, e.g.,:
FIXES: #123456

Tests

Please describe how you tested this change, and include any instructions and/or
commands to reproduce.

Checklist

Before submitting this PR, please make sure:

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have made or will make corresponding changes to any relevant documentation.

[sierraisland]

LGTM