Add support for common SQL-base data loading scripting

[mtakahar]

The framework allows to have multiple create/import, etc. script variations via --ddl-prefix <prefix> option, primarily for accommodating YB specific syntax that would fail on PG.

This mechanism is also useful for testing multiple YB setups, for example:
create.sql - YB, colocated
partitioned.create.sql - YB partitioned tables
postgres.create.sql - no YB dialect for PG
pg_partitioned.create.sql - PG partitioned tables, no YB dialect

Previously, I was able to put INSERT commands in import.sql and use it for populating all the table variants, but the framework started disallowing non-COPY command in import.sql.

Please either allow arbitrary SQL commands in import.sql again or add another script for programatic data loading.