[FIX] Update EOS from config

[zhengy001]

Motivation

Correct EOS as default tokenizer eos_token_id may different from model config eos_token_id

Modifications

Checklist

• Format your code according to the Contributor Guide.
• Add unit tests as outlined in the Contributor Guide.
• Update documentation as needed, including docstrings or example tutorials.

[zhengy001]

Per @remixer-dec, model Mistral-Nemo-Instruct-2407-Q5_K_L.gguf has this issue.

[merrymercy]

Revert the changes in tokenizer_manager.py and sampling_params.py. Move all related changes to schedule_batch.py.
Reason:
The sampling_params is used to handle per-request configs, but this eos_token_ids form hf_config is the same for all requests, so we do not need to attach it to sampling_params.
To reduce the overhead, we should pre-process the hf-config and add some more conditions in check_finished

[zhengy001]

Add an input argument for check_finished as Req cannot acquire model info.