Revise workflow to omit incorrect "daily" and "epiweekly" language.

[damonbayer]

When we were only fitting "e" models, the outputs were either "daily" or "epiweekly," Now that we fit "h" and "e" models, outputs can be mixed, and "h" models are always epiweekly, despite not being aggregated during post processing.

[dylanhmorris]

Could do this in the same PR as #342?

[damonbayer]

As of now, we are producing up to 3 files per models:

• daily_samples.parquet
• epiweekly_samples.parquet
• epiweekly_with_epiweekly_other_samples.parquet

As discussed above, these file names don't really make sense, as the daily and epiweekly labels do not apply to the entire file. This makes it difficult to tell what each of these files contain or if they are redundant.

I propose to, instead, produce a single samples.parquet file, containing all of the samples that were contained in the previous 3 files. We can add three new columns to this file: resolution, ("daily" or "epiweekly"), aggregated_numerator (TRUE or FALSE), and aggregated_denominator (TRUE or FALSE).

Most variables will have NA aggregated_denominator, as they are not computed as fractions.

Do you foresee any issues with this proposal @dylanhmorris?

[dylanhmorris]

This sounds good to me. Thanks @damonbayer!