Parse ARRAY_AGG for Bigquery and Snowflake

[SuperBo]

There are subtle differences of ARRAY_AGG in SQL dialects:

1. https://jakewheat.github.io/sql-overview/sql-2016-foundation-grammar.html#array-aggregate-function
2. https://docs.snowflake.com/en/sql-reference/functions/array_agg.html
3. https://cloud.google.com/bigquery/docs/reference/standard-sql/aggregate_functions
4. https://www.postgresqltutorial.com/postgresql-aggregate-functions/postgresql-array_agg/

This the very first attempt to support correct ARRAY_AGG function while parsing BigQuery and snowflake.

[alamb]

@AugustoFKL do you have time to review this PR?

[coveralls]

Pull Request Test Coverage Report for Build 3435201483

• 58 of 69 (84.06%) changed or added relevant lines in 7 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.01%) to 86.164%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/dialect/mod.rs	1	2	50.0%
src/dialect/snowflake.rs	1	2	50.0%
src/ast/mod.rs	14	17	82.35%
src/parser.rs	28	34	82.35%

Totals
Change from base Build 3413823174:	-0.01%
Covered Lines:	11627
Relevant Lines:	13494

---

💛 - Coveralls

[AugustoFKL]

@SuperBo please fix the linter warnings

[AugustoFKL]

@SuperBo any reason why you did the implementation for all dialects the same time?

You didn't complete none, but added parts of different ones. Got me a little confused

[SuperBo]

I changed this. You can review again

[AugustoFKL]

The comment is confusing.

The comment make it sounds like if any of both is valid, it will return true.

Could be something like

/// Returns true if the dialect supports ARRAY_AGG() [WITHIN GROUP (ORDER BY)] expressions.
///
/// Otherwise, the dialect should expect an `ORDER BY` without the `WITHIN GROUP` clause, e.g. `ANSI` [(1)].
/// [(1)]: https://jakewheat.github.io/sql-overview/sql-2016-foundation-grammar.html#array-aggregate-function

Having the function here is not my usual approach, as it will require maintenance over other dialects individually, but I don't think it's wrong.

[AugustoFKL]

Also, @SuperBo, one thing:

• Generic dialect should be the most permissive one so, if you actually can have both structures in a query, generic dialect should allow both. But considering my previous comment, I'd allow the most common one, maybe?

@alamb what do you think? This is a divergence from dialects. I'd go for the one with more dialects, so the generic is as generic as possible.

[SuperBo]

I think you are right, the ANSI one should be default option.

[SuperBo]

@AugustoFKL, are we good to go with this or should we change to more generics behavior?

[AugustoFKL]

LGTM, sorry the delay

@alamb ready for your eyes 👀

[alamb]

Looks good to me -- thank you @SuperBo and @AugustoFKL