feat: experiments logging improvements

[pd93]

Closes #1979

Refactors the experiments package slightly so that experiments only need to be specified in one place now. Also moved the print functions to the logger to avoid importing anything in the experiments package. This stops circular imports.

This way of doing things allows us to easily iterate over all experiments which means I was able to easily add a function that validates all experiments. If an experiment is inactive (has no enabled values) and a user tries to specify a value, a warning will now be output.

Finally, I've consolidated how we read any Task-namespaced environment variables into the new env.GetTaskVar function that will always prefix TASK_ to any variables read. The logger now only reads these variables once instead of every time a call to print is made.

[andreynering]

👏

[pd93]

@andreynering Have added your suggestion and also added some tests for the experiments and did a little bit more refactoring. We'll now get warnings in two cases:

1. A user attempted to enable an experiment with no valid values

   ❯ TASK_X_ANY_VARIABLES=1 taskd task: Experiment "ANY_VARIABLES" is inactive and cannot be enabled 

2. A user attempted to enable an experiment with a value that is not in the allowed values

   ❯ TASK_X_MAP_VARIABLES=99 taskd task: Experiment "MAP_VARIABLES" has an invalid value "99" (allowed values: 1, 2)