Add max-batch-size to benchmark_throughput.py

[dllehr-amd]

Currently hf-max-batch-size exists to allow hf runs to specify the maximum batch size to use. The vLLM Engine will take in a max batch size, in the form of max_num_seqs. Open this capability to users so we can attempt to change the batch size externally.