Random sampling for T5 text generation

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Right now, the ORT only supports beam search and greedy search, but not random sampling for T5. It would be great to implement random sampling.

System information

• ONNX Runtime version (you are using):
  ORT 1.12

Describe the solution you'd like
A clear and concise description of what you want to happen.

The beam search and greedy search nodes are defined here: https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/python/tools/transformers/convert_generation.py#L889
An implementation of top-k and top-p random sampling for T5 text generation would be great. The corresponding huggingface implementation is here: https://github.com/huggingface/transformers/blob/main/src/transformers/generation_utils.py#L1796

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

Start in onnxruntime/python/tools/transformers/convert_generation.py around the beam search and greedy search nodes at line 889, then compare the Hugging Face generation_utils.py implementation around line 1796. The work is done when T5 text generation supports top-k and top-p random sampling in addition to the existing search modes.