ENH: Add experiment tracking with mlflow as an extra

[NickleDave]

• pip install vak[mlflow] should add this optional dependency and then allow experiment tracking
• we should include at a bare minimum the path to metadata json files from datasets so that one can programatically relate metadata to experiments / runs

I don't want to invest a lot right now in supporting every feature of mlflow but it's clear to me we need some sort of experiment tracking (as @achabotl wisely pointed out many moons ago).

Prefer going with mlflow for now since it's the one I'm most familiar with and I would rather just adopt something that's widely used in industry, although sth more lightweight and application-agnostic like sacred or sumatra might be worth thinking about

[NickleDave]

Just realized lightning has an mlflow logger
https://lightning.ai/docs/pytorch/stable/api/lightning.pytorch.loggers.mlflow.html#mlflow-logger
so we can perhaps somewhat painlessly integrate that way -- pass in experiment name, run name, etc