Support the MLFlfow data tracking in MONAI

[binliunls]

Is your feature request related to a problem? Please describe. The experiment management tool MLFlow released version 2.4.0 last week, which introduced data tracking APIs. With these APIs, users can record some info about the dataset used during training, inference and evalution. This is a very useful feature to reproduce experiments using different subsets of the original dataset. And it would be better if MONAI can also support these APIs in the experiment management part.

Describe the solution you'd like Add the data tracking APIs to the MLFlowHandler API.

Additional context Please note these warnings:

1. Most data tracking APIs are experimental and may change in the future.
2. The API web page doesn't show any support for PyTorch dataset.