Make ref table function way simpler and more efficient

[ehinman]

I realized I initially built the reference table function to match dataRetrieval, thinking that there was a reason we couldn't reuse the get_ogc_data() function, but I think this is inefficient and incorrect. Swapped out the repeated code for get_ogc_data and changed the "id" column to the singular subject name of the reference table, e.g. "site-type-codes" has the "id" column changed to "site_type_code" (the rest of the columns swap underscores for dashes, so keeping consistent).

I also learned that if there aren't any geometries present in the dataframe, geopandas will simply return a pandas DataFrame. This avoids the issue I created in the last MR where I thought I had to specify "no geopandas" to get back a regular dataframe, and the logger was yelling about geopandas not being installed due to my oversight.

The only risk is that URLs contain the argument skipGeometry=FALSE, like this:
https://api.waterdata.usgs.gov/ogcapi/v0/collections/time-zone-codes/items?skipGeometry=False&limit=50000

However, it doesn't seem to actually affect what comes back.

[ehinman]

@thodson, @jzemmels, can one of you take a look at this PR this week?

[jzemmels]

Changes look good. Just one minor typo in the example and a question about dependencies. Otherwise approved.

[jzemmels]

Suggested change

	... collection="parameter-codes
	... collection="parameter-codes"

[jzemmels]

Are these new dependencies from the changes made in this PR?
I don't see docutils or mapclassify used explicitly in the package files anywhere (from a CTRL + ALT + F through the package files)