values probably not columns

i think i'd prefer like:

def inject_additional_insert_values(**kwargs):
    return kwargs

and then e.g.

conn.execute(
    self._run_tags_table.insert(),  # pylint: disable=no-value-for-parameter
    [
        **self.inject_additional_insert_values(
            run_id=pipeline_run.run_id,
            key=k,
            value=v,
        )
        for k, v in pipeline_run.tags.items()
    ],
)

i think i'd prefer a more specific name here but not sure what. it'd be nice to make it clear this was for reads/queries and that it is adding clauses to queries. with_extra_select_clauses_for_table() is terrible but.

if my understanding is correct that concrete implementations of apply_additional_filters and additional_insert_columns will then look up the table name on the table class and then switch on the table name, i don't think i quite understand what the win of passing the table class around is vs. just passing around its name, or the key _bulk_actions_table etc

i think the implementation in the base class should just be:

def apply_additional_filters(self, select: Select, _table_class: str) -> Select:
    return query

interesting, this might be enough -- at least as long as we don't have association tables

i don't think this is necessary

realistically i think all these joins need to be parametrized also, which means that we need some kind of per-query functionality, not just per-table. i think we can perhaps do this in a pluggable way if we introduce some kind of QueryProvider

why?