[XPU] Add TP broadcast after sampling in XPU model runner.

[Jiajun-Ji]

Motivation

💡 If this PR is a Cherry Pick, the PR title needs to follow the format by adding the [Cherry-Pick] label at the very beginning and appending the original PR ID at the end. For example, [Cherry-Pick][CI] Add check trigger and logic(#5191)

💡 如若此PR是Cherry Pick，PR标题需遵循格式，在最开始加上[Cherry-Pick]标签，以及最后面加上原PR ID，例如[Cherry-Pick][CI] Add check trigger and logic(#5191)

In TP, each rank may produce different sampling results due to independent random sampling
Added a broadcast operation after sampling in the XPU model runner to synchronize the sampled tokens from rank 0.

Modifications

Usage or Command

Accuracy Tests

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[EmmonsCurse]

/skip-ci ci_iluvatar
/skip-ci ci_hpu
/skip-ci build_gpu

[Copilot]

Pull request overview

该 PR 旨在解决 XPU + Tensor Parallel（TP）场景下各 rank 独立采样导致的生成结果不一致问题，通过在采样后增加 TP 组内广播，使各 TP rank 以 rank0 的采样结果为准，从而保证跨 rank 的一致性。

Changes:

• 在非投机解码路径中，采样后对 sampled_token_ids 进行 TP 组内广播同步。
• 在投机解码路径中，采样后对 accept_tokens/accept_num/step_idx/stop_flags 进行 TP 组内广播同步。

[Copilot]

这里 src（root rank）的计算表达式在多处重复使用，后面 speculative 分支也同样重复。建议先用局部变量（例如 tp_src_rank = data_parallel_rank * tensor_parallel_size）保存，再传给 broadcast，避免复制粘贴带来的维护风险。

[Copilot]

speculative 分支里连续多次调用 broadcast，且每次都重复同一个 src 计算逻辑。建议复用同一个 tp_src_rank 变量，并考虑用一个 key 列表循环广播这些张量（accept_tokens/accept_num/step_idx/stop_flags），降低后续新增/修改字段时遗漏的概率。