plpgsql: reuse the sql scan logic

Right now, the [PL/pgSQL scanner](https://github.com/cockroachdb/cockroach/blob/752b00dac3ba9bc406a76ca521841937a3df329b/pkg/sql/scanner/plpgsql_scan.go#L27-L29) duplicates much of the logic in the SQL scanner. This is mostly because PL/pgSQL has its own set of tokens, the IDs of which differ from the SQL tokens due to the order in which they're defined in the `plpgsql.y` file. Postgres shares this logic, and is able to do so because it carefully keeps the token IDs in line. We should consider making a similar change.

Jira issue: CRDB-36297