Unify semicolon and comment handling in nested native queries (#30677)

* Unify semicolon and comment handling in nested native queries

* Use different comment and semicolon removal approach

* Apply suggestions from code review

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>

* Update base and effective type verification for oracle

* Avoid running test with snowflake and bigquery

* Update `breakout-results` and test

1. In `breakout-results` case driver statement was redundant.
Original code had field name and types hardcoded. During further
examination, I've realized name and types are available thanks
to the use of `qp.test/col`, so are just "unhardcoded" :)

2. `card-id-native-source-queries-test` is altered, so instead of
hardcoding native sub-queries, roughly equivalent* mbql query is
compiled and its native form is used. This way I can hopefully avoid
problems with extraction of cloud databases' table names for testing
purposes.

* By roughly equivalent I mean instead of using * for selecting
fields, mbql is compiled selecting fields explicitly. Also limit is added
in some cases, but local testing showed no problems with that.

---------

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>