add mrope fusion op

[shaopeng-666]

What this PR does / why we need it?

Add mrope fusion op for qwen2.5-vl. This mrope operator dosen't support Qwen3-VL currently. Thus could only take affect in qwen2.5-vl

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.11.0rc3
• vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

[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 introduces the MRotaryEmbedding to the vllm-ascend project, along with corresponding tests. The MRotaryEmbedding is integrated into the rotary embedding operations and registered as a custom operator. The changes include modifications to tests/ut/ops/test_rotary_embedding.py, vllm_ascend/ops/rotary_embedding.py, and vllm_ascend/utils.py. I have identified a critical issue related to the instantiation of MRotaryEmbedding in the test suite, where the incorrect class is being used, potentially leading to incorrect behavior during testing.

[gemini-code-assist]

critical: The test case is instantiating MRotaryEmbedding from vllm.model_executor.layers.rotary_embedding instead of vllm_ascend.ops.rotary_embedding.AscendMRotaryEmbedding. This will lead to incorrect behavior during testing, as the test will not be using the Ascend-specific implementation.

To fix this, import AscendMRotaryEmbedding from vllm_ascend.ops.rotary_embedding and use that class to instantiate self.layer.

Suggested change

	self.layer = MRotaryEmbedding(self.head_size,
	self.head_size,
	self.max_position_embeddings,
	base=self.rope_theta,
	is_neox_style=self.is_neox_style,
	dtype=torch.bfloat16,
	mrope_section=self.mrope_section)
	from vllm_ascend.ops.rotary_embedding import AscendMRotaryEmbedding
	self.layer = AscendMRotaryEmbedding(self.head_size,
	self.head_size,
	self.max_position_embeddings,
	base=self.rope_theta,
	is_neox_style=self.is_neox_style,
	dtype=torch.bfloat16,
	mrope_section=self.mrope_section)