TRT-LLM Export Code Cleanup

[oyilmaz-nvidia]

What does this PR do ?

Cleaning all of the unused functions, classes, etc, after the TRT-LLM API integration.

[jiemingz]

get_tensor_parallel_group is not used anymore.

I think this one too:

[jiemingz]

I believe cpu_map_location is duplicated with this

Also it seems to me cpu_map_location and gpu_map_location should be in trt_llm/nemo/nemo.py since thats where they're only used

[jiemingz]

This function is only used by build and refit so it can be removed, eventually though we'll have to add it back as well when build and refit are moved onto the new API

[oyilmaz-nvidia]

Thanks @jiemingz for your review. Updating the code per your comments.