[GTP] Documentation update

[fanshiqing]

What's need to update later:

• Best practices related

  • add MEGATRON_LOGGING_LEVEL=20 if you want to check the GTP prefetch chain
  • PYTORCH_CUDA_ALLOC_CONF expandable_segments:True,graph_capture_record_stream_reuse:True,pinned_max_round_threshold_mb:128
  • Don't set CUDA_DEVICE_MAX_CONNECTIONS for Blackwell+

• Latest perf / scalability on nemotron-next

• Update supporting matrix for now

• Others...