[Main][Ops] Make triton rope support index_selecting from cos_sin_cache (#5450)

### What this PR does / why we need it?

This PR extends original `rope_triton_forward` and
`split_qkv_rmsnorm_rope` to support `cos_sin_cache` && `positions` as
inputs. This fully aligns to vLLM RoPE api interface. Compared with
earlier implementation for RoPE, the benefits are:

1. avoiding pre-computation of `cos` `sin` before model execution, which
helps to remove redundant codes.
2. allowing eagle3 draft model to have different rope parameters with
main model (see #6612 ). This help to recover accept rate && accuracy in
that case.

In addition, this kernel change only introduces very small performance
degradation. Those `index_select` or `chunk` operations are now changed
into simple memory access in triton kernel (For example,
https://github.com/vllm-project/vllm-ascend/pull/5450/changes#diff-a4c2d3071530df193b98f9bf38553874bc4d47571336711f116c26d019cfbb6aR77-R81).

**Highlights**

- **RoPE Cache Unification**: Replaced separate _sin and _cos global
tensors with a unified cos_sin_cache and explicit positions tensor for
Rotary Positional Embeddings (RoPE), streamlining data handling.
- **Triton Kernel Integration**: Updated Triton kernels
(split_qkv_rmsnorm_rope_kernel, _triton_rope) to directly consume the
cos_sin_cache and positions for more efficient and integrated RoPE
calculations.
- **Custom Operation Registration**: Registered `rope_forward_oot` as a
new custom operation, allowing its use in fused compilation passes and
providing a dedicated entry point for the new RoPE implementation.
- **Refactored RoPE Forward Pass**: Modified the rope_forward_oot
function to accept the new cos_sin_cache and positions arguments,
enabling a more flexible and integrated RoPE application within the
system.

### Does this PR introduce _any_ user-facing change?

No.

### How was this patch tested?

- vLLM version: v0.13.0
- vLLM main:
5326c89803

Additional test on Qwen3-235b accuracy:

| Aime2024 | GSM8K | Livecodebench |
| -------- | -------- | -------- |
| 83.33 | 96.26 | 70.23 |

---------

Signed-off-by: Angazenn <supperccell@163.com>

vllm_ascend/spec_decode/eagle_proposer.py

@@ -39,7 +39,6 @@ from vllm_ascend.attention.attention_mask import AttentionMaskBuilder
 from vllm_ascend.attention.attention_v1 import AscendAttentionState
 from vllm_ascend.attention.utils import AscendCommonAttentionMetadata
 from vllm_ascend.compilation.acl_graph import ACLGraphWrapper, update_full_graph_params
-from vllm_ascend.ops.rotary_embedding import update_cos_sin
 from vllm_ascend.ops.triton.spec_decode.utils import prepare_inputs_padded_kernel
 from vllm_ascend.ops.triton.triton_utils import get_vectorcore_num
 from vllm_ascend.utils import enable_sp, lmhead_tp_enable, shared_expert_dp_enabled, vllm_version_is
@@ -299,9 +298,6 @@ class EagleProposer(VllmEagleProposer):
             _,
         ) = self.runner._sync_metadata_across_dp(num_tokens, is_draft_model=True)
 
-        # update global cos, sin
-        update_cos_sin(self._get_positions(num_tokens))
-
         multi_steps_attn_metadata = []
         if not self.use_cuda_graph:
             aclgraph_runtime_mode = CUDAGraphMode.NONE
@@ -471,9 +467,6 @@ class EagleProposer(VllmEagleProposer):
         builder = self.runner.attn_groups[0][0].get_metadata_builder()
         attn_metadata = builder.build(0, common_attn_metadata, self.runner.get_model())
 
-        # update global cos, sin
-        update_cos_sin(self._get_positions(num_input_tokens))
-
         if self.uses_mrope:
             used_update_positions = target_positions[:, last_token_indices]
         else:
@@ -651,9 +644,6 @@ class EagleProposer(VllmEagleProposer):
                 input_ids = self.input_ids[:input_batch_size]
                 inputs_embeds = None
 
-            # update global cos, sin
-            update_cos_sin(self._get_positions(input_batch_size))
-
             # Run the model.
 
             # The lifecycle of `input_ids`, `positions`, `hidden_states` runs through all