[Feature Request] transformers convert_generation.py only support three types： beam_search，greedy_search and sampling?

### Describe the feature request

I have noticed that in https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/python/tools/transformers/convert_generation.py, it only supports three types: beam_search，greedy_search and sampling.

```python
class GenerationType(Enum):
BEAMSEARCH = "beam_search"
GREEDYSEARCH = "greedy_search"
SAMPLING = "sampling"

```

Is there support for other types, such as beam_sample? If not, are there any plans to support it?

### Describe scenario use case

In fairseq\transformers\src\transformers\generation\utils.py, there are 7 types of generation, contrastive_search, greedy_search, sample, beam_search, beam_sample, group_beam_search and constrained_beam_search.

Start by reading onnxruntime/python/tools/transformers/convert_generation.py, especially the GenerationType enum, and compare its supported modes with fairseq/transformers/src/transformers/generation/utils.py. Determine the scope and design needed for additional generation types such as beam_sample; done means the supported behavior and plans are clearly established and the converter handles the agreed types.

Written by the indexing model from the issue text.