Proposals to simplify the workflow

[aaTman]

I think that I am getting more comfortable with finding some nits in the workflow that I think we can update and fix. I would like to have these ready for v1.0; I think they would GREATLY reduce friction for all users.

Case metadata:

1. I don't think event_type should be a requirement for our case metadata. It should be up to the user to identify what their event type is. For example, separating a bunch of cases into different YAMLs.

2. I also think we can remove case_id_number as a requirement.

3. We should create a method to allow users to include any arbitrary sets of metadata in the case metadata, so that (if they want) they can add event_type, case_id_number, or whatever else (e.g. was_i_located_here_at_the_time or which_mode_of_transportation_did_this_impact)

Overarching workflow

1. Importing all of the modules is a nuisance. Instead of from extremeweatherbench import defaults, cases, metrics, derived, evaluate, I think something like import extremeweatherbench as ewb and having each as a distinct submodule would reduce unneeded verbosity. An example here then would be ewb.metrics.MeanAbsoluteError() or ewb.inputs.ERA5(). This change would also reduce risk of users unintentionally overriding modules with variable calls for things we're commonly interacting with, like cases or metrics.

2. Also, to build on 1, we would bury the ExtremeWeatherBench class in evaluate.py away from the user and instead have it directly callable, e.g. ewb.evaluation(...)

3. I have already identified it in No simple way to build a list of targets or forecasts without needing metrics #117 but there should be a way to easily grab the subset target or forecast data without setting up a full blown evaluation.

I have a branch set up for 2 which should handle it.