Add PRIMARY KEY Aggregate support to dataframe API

[mustafasrepo]

Which issue does this PR close?

Closes #.

Rationale for this change

Currently we can rewrite group by expressions according to primary key.
Consider query below

SELECT ts, sn, SUM(amount) as sum1
  FROM sales_global_with_pk
  GROUP BY sn

normally, after aggregation only sn and SUM(amount) as sum1 would be available for output. However, given sn is PRIMARY_KEY, we know that each sn value maps to a fixed ts value (e.g They are functionally depandant). Hence, by treating query above as

SELECT ts, sn, SUM(amount) as sum1
  FROM sales_global_with_pk
  GROUP BY ts, sn

under the hood, we can output ts also at the end. However, this feature was only available with SQL API. See discussion.

This PR brings this support to the dataframe API also. With this PR following dataframe query can be executed given col_id is PRIMARY_KEY.

        // group by contains id column
        let group_expr = vec![col_id.clone()];
        let aggr_expr = vec![];
        let df = df.aggregate(group_expr, aggr_expr)?;

        // expr list contains id, name (e.g projection)
        let expr_list = vec![col_id, col_name];
        let df = df.select(expr_list)?;

What changes are included in this PR?

Are these changes tested?

Yes, most of the changes comes from either plan tests, or .slt test

Are there any user-facing changes?

[alamb]

I plan to review this PR tomorrow if no one beats me to it

[ozankabak]

I reviewed this very carefully (over 6 sittings as you can see in the commit history!) and it is almost ready to merge. It pays a longstanding tech debt where functional dependencies worked in SQL but not in dataframe API.

I asked @mustafasrepo to make one little optional fix and then it will be ready to go.

[ozankabak]

Since this has been around for a week, it doesn't make any potentially disruptive change, pays an old technical debt, and that I have already reviewed it very carefully; I will go ahead and merge this.

In case anything breaks, let us know and we will promptly fix it.