Puma/EE: Scale Geo-DB pool size with thread count
This was broken out of #35170 (closed), as it is a behavioral rather than structural change.
In 1a53abd4 we introduced a change whereby the main database connection pool is scaled according to the number of Puma worker threads. However, we did not apply the same change to Geo database connections when running Puma under EE. We already perform that sort of scaling for Geo when running Sidekiq (previously in the sidekiq initializer; once !22385 (merged) is merged in database_config), so it stands to reason that this should happen under Puma as well.