Fix bugs in SQLScriptScanner with big String literals and PostgreSQL identifiers (as introduced by #7646)

[inponomarev]

"This never happened before, and here we go again": unfortunately #7646 introduced not only proper bug fixes, but some new bugs as well, reproducible on some of our test SQL scripts.

This PR introduces 3 changes:

1. (⚠️ The biggest problem). It seems that Java's regexp "(\\"|[^"])*", though working well with strings of a reasonable size, tends to yield StackOverflowError when literals are big (10k simbols and more). The workaround is the same as with comments: use finite-state machine to match the literal, not the regexp.
2. According to PostgreSQL docs, dollar sign can be a part of an identifier (non-standard behaviour for SQL, as they themselves admit), and it must be handled so that this$is$a$valid$postgreSQL$identifier is not mixed with dollar-quoted strings
3. m.find(offset) && m.start() == offset was suboptimal (we are only interested in cases when the regexp matches exactly starting from the offset and there is no reason in finding it somewhere far away in the script).

Existing tests are not modified, new tests added.

[kiview]

The checkstyle issue breaks CI.

I am not sure, why it is not correctly displayed as failing CI, I'd say our find_gradle_jobs task is not correctly being recognized as failing.

You can run the checkstyle checks locally with ./gradlew :database-commons:spotlessCheck.

[eddumelendez]

Thanks for your contribution, @inponomarev !