[Feature Request] Experiment manager

Let's say I conduct a lot of experiments with parameters and generate more than 10,000 images per week.

Currently, there is no easy way to track such a large number of experiments, and many parameters are missing, making it very difficult to reproduce the results and share them with others.

While the workflow.json operates at the workflow level, much vital information is per run and is currently absent.

The proposed solution is to implement an experiment manager to:

1. Track all information of each run.
2. Rerun experiments with the same parameters.
3. Visualize the results of each run.
4. Share experiment runs.
5. Automatically deploy each run to the cloud, etc.

It's kind of like MLflow for machine learning.