Add a new flowmachine query class that can handle "`MajorityLocation` joined to `LocationVisits`"

[jc-harrison]

We would like to be able to define a query that returns a specified reference location (we are particularly interested in MajorityLocation for each subscriber, only if they visited that location on more than 'threshold' days. This could be achieved with something like:

SELECT subscriber, location
FROM {reference_locations}
INNER JOIN {subscriber_locations_subset}
USING (subscriber, location)

where the parameters could be reference_locations=MajorityLocation(...) and subscriber_locations_subset=LocationVisits(...).numeric_subset(low=threshold, high=max_days). Alternatively we could use INTERSECT or INTERSECT ALL instead of an inner join - either would treat duplicates differently from the join, and the choice would presumably have performance implications.

This whole step could be achieved using existing methods, using:

reference_locations_query.join(
    location_visits_query.numeric_subset(low=threshold, high=max_days)),
    how="inner",
    left_on=('subscriber', <location_id_columns>),
    right_on=('subscriber', <location_id_columns>),

but this is not ideal because:

• we'd like to use this query as an input to CoalescedLocation (and quite possibly other queries), which would expect to recognise it as a subscriber location query (i.e. having a spatial_unit attribute), which is not the case if the query is a generic Join query,
• exposing generic Join through the API wouldn't help, for much the same reason - only specific joins would be suitable parameters for the queries that could consume this one, so we'd need a specific marshmallow schema for those special-case joins, which would mirror the flowmachine query we're proposing to add here.

[jc-harrison]

The most pragmatic option here might be to add a new SubscriberLocationJoin subclass of Join, that internally sets the join columns and adds a spatial_unit attribute.