[bugfix] the hccl_buffsize config for EP

[kiscad]

What this PR does / why we need it?

The hccl_buffsize for EP process group is set by the _DEFAULT_BUFFER_SIZE, which is not enough for fused_moe kernel, especially for large max-num-batched-tokens. So we need to calculate the hccl_buffer for EP from the vllm_config.

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.11.2
• vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.2

[github-actions]

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

[gemini-code-assist]

Code Review

This pull request fixes an issue with hccl_buffsize configuration by ensuring the HCCL_BUFFSIZE environment variable is respected when creating HCCL process group options. The changes look good, but I have a couple of suggestions. First, the logic for parsing the buffer size from the environment variable should be improved to handle non-positive values to prevent potential issues. Second, since the special handling for "mc2" communication groups was removed, adding a specific test case for this scenario would increase confidence that this change does not introduce any regressions.