Advanced class-based `conversions=` mechanism

The conversions= parameter works like this at the moment: https://sqlite-utils.datasette.io/en/3.23/python-api.html#converting-column-values-using-sql-functions

db["places"].insert(
    {"name": "Wales", "geometry": wkt},
    conversions={"geometry": "GeomFromText(?, 4326)"},
)

This proposal is to support values in that dictionary that are objects, not strings, which can represent more complex conversions - spun out from #399.

New proposed mechanism:

from sqlite_utils.utils import LongitudeLatitude

db["places"].insert(
    {
        "name": "London",
        "point": (-0.118092, 51.509865)
    },
    conversions={"point": LongitudeLatitude},
)

Here LongitudeLatitude is a magical value which does TWO things: it sets up the GeomFromText(?, 4326) SQL function, and it handles converting the (51.509865, -0.118092) tuple into a POINT({} {}) string.

This would involve a change to the conversions= contract - where it usually expects a SQL string fragment, but it can also take an object which combines that SQL string fragment with a Python conversion function.

Best of all... this resolves the lat, lon v.s. lon, lat dilemma because you can use from sqlite_utils.utils import LongitudeLatitude OR from sqlite_utils.utils import LatitudeLongitude depending on which you prefer!

Originally posted by @simonw in https://github.com/simonw/sqlite-utils/issues/399#issuecomment-1030739566

Start at the db["places"].insert(..., conversions=...) entry point and read the existing conversions contract in the linked Python API documentation. Then inspect the proposed helpers from sqlite_utils.utils; done means conversions can accept an object combining a SQL fragment with Python value conversion, including the LongitudeLatitude and LatitudeLongitude variants.