Minor fixes for MBridge nemotron deployment

[athitten]

Adds minor fixes required for deployment of nemotron.

1. attention_backend from run_config.yaml in the MBridge ckpt can sometimes be read as str type, so adds code to in nemo_deploy/llm/inference/inference_base.py to convert it to AttnBackend enum as expected by MCore.
2. Minor fix to check if tp_comm_overlap_cfg exists in the config and then check if its not none, as it can otherwise lead to AttributeError if tp_comm_overlap_cfg is not in the config.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[athitten]

/ok to test b3566b4

[athitten]

/ok to test b3566b4