Add a config mechanism

[suvayu]

• config.py: add config object
• scripts/config.toml: example config

Features

• The configs are namespaced such that separate sections can be read as separate config objects.
• The config objects are also easily created from a dictionary, so that it's easy to use programmatically.
• There are no mandatory fields; there are defaults when fields are missing, some are None because I didn't know what would be most appropriate.
• The config also does type validation in a very limited capacity. Changing the types arbitrarily will most likely fail (e.g. nested types), but plain old data types, and basic containers like list (including multi-dimensional), dict as long as they are flat will validate correctly.

How to use?

@HannoSpreeuw, I wonder if you have any preference how to use the config object. Typically I would adjust the code so that where relevant you pass a config object.

[HannoSpreeuw]

Let me get into this in one of the coming days.

This is about fixing issue #74 , right?

[suvayu]

Yes, that's right

[HannoSpreeuw]

I can't immediately grasp the idea behind the _Validate class and the validate_nested function.

[suvayu]

I can't immediately grasp the idea behind the _Validate class and the validate_nested function.

They are my implementation to do type validation without adding additional dependencies. As I mentioned in the PR intro, for simple flat data structures, it should work. It's somewhat future proof.

Basically it prevents the user from providing incorrect data types in the config.

[HannoSpreeuw]

Allright, aside from a few adjustments - removing sep and appropriate default values - we can move forward after fixing the build problem: ModuleNotFoundError: No module named 'tomllib'.

[HannoSpreeuw]

The unit tests may pass since the global variables are still used.