[recipe] feat: add TRTLLM FP8-E2E recipe for Qwen3-30B-A3B DAPO on GB200

[Superjomn]

The result should roughly align with the [vllm recipe's experiment](https://verl.readthedocs.io/en/latest/low_precision/fp8.html#experiments-and-results):

[gemini-code-assist]

Code Review

This pull request introduces a new end-to-end FP8 training recipe for Qwen3-30B-A3B using Megatron and TRT-LLM, along with updated documentation in the README. Feedback on the new shell script includes wrapping the logger list in double quotes to prevent bash globbing issues and addressing an unused variable (rollout_token_veto_threshold) that is defined but not mapped to the algorithm configuration.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[sophiayyya]

Hi @Superjomn It seems that the entropy is higher than entropy in https://verl.readthedocs.io/en/latest/low_precision/fp8.html#qwen3-30b-a3b-moe-model. Have you applied TIS? And it would be better to have a comparison between fp8 and bf16.