[Feature]: Decouple max_batch_size from cuda_graph_batch_sizes

[MrGeva]

🚀 The feature, motivation and pitch

Currently we always capture cuda graphs up to the highest value of cuda_graph_batch_sizes.
if the user sets a higher max_batch_size it will fail on assertion in CapturedGraph because input batch size dim will be higher than the maximal cuda_graph_batch_sizes value. need to capture up to the max value of cuda_graph_batch_sizes and fallback to non-captured graph when higher batch size occurs.

Alternatives

No response

Additional context

No response

Before submitting a new issue...

• Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.