Reviewing database columns

[dch0ph]

As a task from #5027, we should review the database columns.

I don't know an objective way to do this; it would require some skill and a lot of work to quantitatively assess effects on database performance. But here are some quick thoughts:

• We have previously considered whether foot, bicycle and horse should be relegated to hstore. Although used frequently (e.g. foot 9.9m), they are not important for "primary selection" and only really used in conjunction with other access tags, such as motor_vehicle which are in hstore. Having all the mode-specific tags in hstore would allow us to simplify the arguments to carto_highway_int_access and only pull out the access tags as required for a specific highway type.
• lock (14.7k) seems a prime candidate for relegation. Not even used in the selection for water-lines-text.
• route (1.2m) only seems to be used for ferry routes.
• military (181k) and aerialway (206k), despite being top-level tags hardly seem to justify their own columns, but are used in multiple queries. Leave as is?
• religion (1.8m) is not used in selection. There doesn't seem a reason for this to have its own column when other tags, such as sport (2.8m) and location (3.2m) are pulled from hstore in the amenity query.