[fp8]: Read detected local world size environment variable

[tandede]

📌 Checklist before creating the PR

• I have created an issue for this PR for traceability
• The title follows the standard format: [doc/gemini/tensor/...]: A concise description
• I have installed pre-commit: pip install pre-commit && pre-commit install

🚨 Issue number

Fixed #6437

📝 What does this PR do?

process_group_is_intranode() previously detected the first available local-world-size environment variable but always read LOCAL_WORLD_SIZE. Jobs launched with OpenMPI or Slurm therefore raised KeyError before FP8 collectives could select the intranode path.

This change reads the detected variable and stops at the first match, preserving the existing priority order: torchrun, OpenMPI, then Slurm.

Regression tests cover:

• torchrun, OpenMPI, and Slurm environment variables independently;
• intranode and cross-node process groups;
• deterministic priority when several variables are present.

Validation:

• pytest -q --confcutdir=tests/test_fp8 tests/test_fp8/test_fp8_utils.py — 7 passed
• pre-commit run --files colossalai/quantization/fp8.py tests/test_fp8/test_fp8_utils.py — all hooks passed

The repository-level test fixture invokes GPU accelerator cache hooks during collection, so the full GPU suite is left to CI; the new regression itself is CPU-safe.

💥 Checklist before requesting a review

• I have linked my PR to an issue
• My issue clearly describes the problem and includes a minimal reproduction
• I have performed a self-review of my code
• I have added thorough tests
• I have added docstrings for all the functions/methods I implemented (no new functions or methods)

⭐️ Do you enjoy contributing to Colossal-AI?

• 🌝 Yes, I do.
• 🌚 No, I don't.