Improve datatypes: CHARACTER VARYING for ANSI, and Postgres and also TIMESTAMP AT TIME ZONE for Postgres

[WittierDinosaur]

Brief summary of the change made

Added CHARACTER VARYING as parsable data types to ANSI
Added CHARACTER VARYING and TIMESTAMP AT TIME ZONE "MY_TIMEZONE" as parsable data types to Postgres

Fixes #1331
Fixes #733

Are there any other side effects of this change that we should be aware of?

No

Pull Request checklist

Tests added for both of the bugs' examples.

[tunetheweb]

Added to ANSI, as I don't think having these available is breaking to other languages.

I'm not so sure about that decision to be honest. BigQuery, TeraData, and T-SQL already have their own definition of DataTypeSegment. Should we not just add one for Postgres unless there's a good reason to add this to ANSI?

[WittierDinosaur]

Can we make a deal? Character Varying is an ANSI data type. TIMESTAMP AT TIME ZONE is dialect-specific. But, ANSI is also missing Data Types. For example, CHARACTER LARGE OBJECT. This PR fixes live issues. I agree that TIMESTAMP AT TIME ZONE should be Postgres-only. But I think the types logic needs an overhaul anyway, it's too generic right now. So can we merge this, then look at setting the types right across dialects?

[tunetheweb]

Can we make a deal? Character Varying is an ANSI data type. TIMESTAMP AT TIME ZONE is dialect-specific. But, ANSI is also missing Data Types. For example, CHARACTER LARGE OBJECT. This PR fixes live issues. I agree that TIMESTAMP AT TIME ZONE should be Postgres-only. But I think the types logic needs an overhaul anyway, it's too generic right now. So can we merge this, then look at setting the types right across dialects?

What's your thinking here for not splitting ANSI and Postgres now? While I agree ANSI is not complete, that's not a reason to knowingly add an incorrect one to that, when we're going to have to split our Postgres at some point anyway. Especially when lots of other dialects already have their own implementations of this section. If it required a whole lot of reengineering I could maybe see your point but since it simply requires a copy and paste of the ANSI dialog to the Postgres dialect and then removing the "TIMESTAMP AT TIME ZONE" that your added from ANSI, then why not just do it in this PR?

[WittierDinosaur]

Okay, fair point. Changed.

[tunetheweb]

Thanks for humouring me!

Just one small thing I spotted and then think we're good to merge.