Implement OpenAI Whisper Model

[MahmoudAshraf97]

This PR aims to enable exporting HF Whisper implementation to TRT-LLM checkpoint.
Although TRT already has a conversion script for both HF and original implementation, but implementing it here enables users to use modelopt for optimization and quantization techniques that aren't available yet in TRT-LLM

There are some sharp edges in this implementation, mostly because this is the second Enc-Dec model to be implemented after T5 so they are handled as edge cases, perhaps having a generalized Enc-Dec path can be better, also there might be some code duplication between whisper and T5 in TRT-LLM config creation that can be unified

[yuekaizhang]

@MahmoudAshraf97 Hi, thanks for your effort. I would take this PR into our internal gitlab (We don't have CI running in Github). Also, we would add your name into the co-author list and credit your work on the release notes for whisper fp8 quant support.