Make universal search weighting harder to manipulate

[kueda]

Currently we generally use number of observations associated with a record to weight universal search results, e.g. if a project, taxon, and project all have the name "California" and 10, 10000, and 100 observations respectively, the taxon will appear first. This mostly works, but unfortunately collection projects can be configured to have a ton of observations just by having some very coarse filters, e.g. all animals, or by making an umbrella project that contains several other projects with lots of observations. I'm fairly sure this is why https://www.inaturalist.org/projects/city-nature-challange-2019 is the top results when you search for "city nature challenge 2019" even though it's misspelled. This problem could also be exploited to pollute search results in other ways, e.g. making a giant umbrella project that's actually spam.

I don't think we need to ditch the observations count method entirely, but it shouldn't be what we use to weight projects. Some other ideas:

1. Use the observations count of the project admin, or a sum of the observation counts of all project admins
2. Use the sum of the observation counts of all project curators
3. Use sum of observation counts of all project members