Add pre-config hook

[steve-chavez]

Closes #2703.

Allows using the in-database configuration without SUPERUSER. You can have a function that configures postgREST like so:

create or replace function postgrest.pre_config()
returns void as $$
begin
  if current_user = 'other_authenticator' then
    perform
    , set_config('pgrst.db_schemas', 'schema1, schema2', true)
    , set_config('pgrst.jwt_role_claim_key', '."other"."pre_config_role"', true)
    , set_config('pgrst.db_anon_role', 'pre_config_role', true)
    , set_config('pgrst.db_tx_end', 'rollback-allow-override', true);
  else
    null;
  end if;
end $$ language plpgsql;

The function name and schema is arbitrary. It should be kept separate from public APIs(not unsafe to expose it bc it doesn't return anything, but for maintenance).

A single function can be used for different postgREST instances. Each instance can use a different connection role and different settings can be applied(like the above example with the condition on current_user).

[steve-chavez]

Just tried this for reproducing #2791. No dice.

It happens before the schema cache reload on Workers.hs, which makes sense(schema cache needs the db-schema config).

We need a way to inject a function into the schema cache for reproducing #2791. This would also be useful for inducing a Hasql error and increase code coverage(currently we cannot cover the InternalErrorCode00/PGRSTX00 errors). This should be an internal only config parameter bc there's no feature that needs this.