Check hipblaslt tuned performance vs default

[causten]

HipBlasLt created a document on how to tune your own hipblas calls. https://github.com/ROCm/hipBLASLt/blob/develop/docs/how-to-use-hipblaslt-offline-tuning.rst

Generate a script that can capture the calls from an onnx model, run hitblaslt-bench before and after then rerun the original model. The goal would be to identify if a tuning request for a given model needs to be requested.

[ahsan-ca]

I had a discussion with the hipBLASLt team and the way we do our tuning is similar to how it is outlined in the document, with the only difference being in the number of iterations. I am doing analysis of number of warmup/tuning iterations on the performance. I think we can close this issue since we do not need to write a script to capture hipblaslt-bench commands.
I will open another issue to track the result of analysis of iterations on performance for hipBLASLt.

[ahsan-ca]

Additional work followed up in the issue: #3804