The cyclomatic complexity in `describe_screens.py` is too high

[phicks22]

We need to condense and optimize the collection of metadata from the IDR API while looping through image ids.

[gwaybio]

I still think that grabbing these details from https://github.com/IDR/idr-metadata will dramatically reduce complexity. You might still need to grab image count per well somehow elsewhere, but I think it will eliminate the need to loop through image ids.

[gwaybio]

Now that we have example .parquet files (added in #7 ) - Is it worth comparing these to the IDR metadata files?

If they contain the same information, then it argues that we don't need to loop through image IDs for the purpose of collecting descriptive statistics. (we may still need to iterate through images for other reasons)