Add optional arguments to filter forecast outputs in `hub_to_scorable_quantiles`

[sbidari]

Allows applying filters before collecting which will help avoid loading large data into memory especially for large hubs like FluSight

This will be useful in https://github.com/cdcent/cfa-forecast-hub-internal-reports/issues/104

[dylanhmorris]

I think we should consider deprecating or at least rewriting the internals of hub_to_scorable_quantiles etc now that hubEvals (as of March 9) has user-exposed functions for converting hubverse data to scoringutils tables for pretty much everything we care about.

[sbidari]

I don't think hubEvals exposes anything similar to the scorable table (joined table with both predicted and observed data) which we use to generate predicted observed plots

[dylanhmorris]

I think it does now: https://hubverse-org.github.io/hubEvals/reference/transform_quantile_model_out.html