[MRV2][Misc] Align MRV2's public code logic with upstream (#14100)

### What this PR does / why we need it?
- Refactored Model Runner Inputs: Updated the prepare_inputs method in
vllm_ascend/worker/v2/model_runner.py to utilize BatchReqState for
managing request-related data, streamlining input preparation.
- Improved Input Data Handling: Replaced manual mapping and sorting
logic with pre-computed values from BatchReqState, improving efficiency
and reducing redundant calculations during the input preparation phase.

### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Tested with CI

- vLLM version: v0.27.1
- vLLM main:
https://github.com/vllm-project/vllm/commit/ba07e4a48fc951300d97eb506217dd530583dea3

Signed-off-by: zouzy <zouzongyu@huawei.com>