Aeromancy configuration should live in pyproject.toml

[dmcc]

There's lots of Aeromancy customization which needs to live somewhere. We currently do it by baking in command line arguments to the pdm go script, but much could be specified more cleanly in pyproject.toml. This includes --extra-docker-run-args, --extra-env-var, etc.

And potential new features like:

• Support Podman instead of Docker #15
• Support MLFlow as an experiment tracker #16
• Store default W&B entity/project (?)

Aeromancy already uses the tomli package (which seems to be API-compatible with newer stdlib tomllib), so that might be an easy way to get this in.

One question is how to combine TOML configuration with command-line flags. In general, I think they should be additive, though for flags that set command line arguments (e.g., --extra-docker-run-args), we might want to ensure that they're only set in one place and raise an error otherwise.