Support evaluation of Megatron-Bridge checkpoints

[athitten]

Support evaluation of Megatron-Bridge) checkpoints: verify/add Open AI API compatibility in the deployment and check eval accuracy scores.

[athitten]

Being worked on this week.

[athitten]

Generation benchmarks gave the expected accuracy on GSM8K, MMLU Llama 3.18B. LogProb benchmarks however are showing a lower accuracy than baseline with arc_challenge, winogrande on Llama 3.2 1B Instruct and Llama 3.1 8B. Investigating this issue.

[athitten]

PR for Export-Deploy to enable this feature: NVIDIA-NeMo/Export-Deploy#445

[athitten]

Docs need to be added, reopening the issue.

[athitten]

Docs PR: #388

[athitten]

QA please use the docs here: https://github.com/NVIDIA-NeMo/Evaluator/blob/athitten/mbridge-eval-docs/docs/nemo-fw/evaluation-doc.md. Thanks

[athitten]

QA Please use the docs here for this feature: https://github.com/NVIDIA-NeMo/Evaluator/blob/main/docs/nemo-fw/evaluation-mbridge.md