[Kernel] add triton kernels for sampling

[swy20190]

What this PR does / why we need it?

Replace pyorch implement of sampling with triton kernels

Does this PR introduce any user-facing change?

No

• 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 replaces several PyTorch-based sampling functions with Triton kernels, which should improve performance. The new kernels for greedy and random rejection sampling, batch expansion, and recovered token sampling appear to correctly replicate the logic of the original PyTorch functions. I've found a critical syntax error in the new Triton-related imports that will prevent the code from running. I've left a specific comment with a suggested fix. Additionally, it appears the unit tests in tests/ut/sample/test_rejection_sampler.py have not been updated to reflect the removal of the PyTorch functions and the introduction of the Triton kernels. These tests will fail and should be updated to validate the new implementations.

[gemini-code-assist]

This import statement contains a syntax error. The from ... import ... syntax does not allow dot notation like triton.language in the imported names. To fix this and follow the common pattern in vLLM for handling the optional Triton dependency, you should import tl directly from vllm.triton_utils and combine the imports into a single line for conciseness.

from vllm.triton_utils import HAS_TRITON, triton, tl