override seed types

[drewbanin]

The following config will force the create table statement for the country_codes.csv file to use a text type for the id field.

seeds:
    my_project:
        country_codes:
            column_types:
                id: text
                other_field: numeric(12,4)

I don't love the idea of adding another arbitrary config to the seeds block, but I'm unsure where else we can implement this.

I know some folks also want this functionality for models too, but we don't have a good way of overriding column types for create table as (...) statements. @cmcarthur really keen to hear your thoughts about all of this.

Regarding the code in this PR: I think a lot of it should be ripped out of the adapters. If possible, I'd like to make a seed materialization which encodes this logic, then execute the seed nodes as we would any other node. I don't love the amount of business logic in the SeedRunner class

TODO:

• Extend tests to also run for BQ, Snowflake

[cmcarthur]

I think this is fine. I agree that this code should be ripped out of adapters. This PR is sort of confusing to read because of adapter inheritance and because of this being done in slightly different ways on BQ vs PG/Snowflake/Redshift (_agate_to_schema in one place, inline in the other). But I don't want to commit to that type of rewrite right now, so this is approved

[drewbanin]

Just added a couple more tests for bq, snowflake. Will merge when green again

[drewbanin]

Closes: #703, #690