Alaska and Virginia changes: #2136

Regarding statistical gore: US Census does provide attributes to indicate if a certain geography is "real" or "statistical" via the FUNCSTAT code. Only 82 of these across 3.2k features (2.5%).

• F - Fictitious Census Entity (43) places like Baltimore county (fake) versus Balitmore city (real) and almost all the Virginia city-county places, what WOF calls coterminous or placetype alt. It'd require some sleuthing to see if WOF matches US Census in which "primary" feature is "real" versus "fake", and WOF always views the locality as primary so this might not match always anyhow? But they are real places except for the hierarchy padding bit, we could set these as statistical gore and possibly hierarchy_label = 0
• N - Nonfunctioning Legal Entity (28) places like in Connecticut, Rhode Island, and most of Massachusetts, plus few US territory places like Swains Island. These ones should be marked statistical gore in WOF. Seems like they should not show up in hierarchy label.
• S - Statistical Entity (11) the Alaska census areas. These ones should be marked statistical gore in WOF. Should actually show up as hierarchy label, as there isn't anything else around.

See also:

• https://www.census.gov/library/reference/code-lists/functional-status-codes.html

The new counties mean the localadmin in Connecticut need new uscensus:geoid concordance values, and also populations since the earlier joins on outdated concordance didn't yield a match. This is 169 features.

I also noticed a handful of localadmin outside of Connecticut that didn't have a good join, like Amherst in Mass. In some rare cases it seems like we had the same geoid on the locality as the localadmin when the localadmin should have a unique geoid. This also affected the population join and should be fixed. This affects 17 features in 9 states.

If we apply the same statistical gore lens to all localadmin across the entire USA, the trend is 4.6k out of 23.4k (20%). Spot checking these do all seem to be statistical placeholders with lower risk for false categorizing, even for F entities. To explain how we got there in the future, we should always be importing uscensus:funcstat now and in the future (21k out of 23.4k or 90%).

Connects with whosonfirst-data/whosonfirst-data-admin-us#207