Suggest columns after GROUP BY

[Kynde]

What

Suggest columns after GROUP BY, the same way pgcli already does for ORDER BY.

Why

The completion dispatch in suggest_based_on_last_token offers column suggestions for SELECT, WHERE, HAVING, ORDER BY and DISTINCT, but GROUP BY was missing from that keyword tuple. As a result, a query like:

SELECT mfa_method, count(*) FROM companies GROUP BY <tab>

fell through to the generic keyword handling and ultimately re-derived its suggestions from the FROM clause — offering tables/schemas instead of the table's columns. ORDER BY, HAVING and WHERE in the same position all correctly suggest columns.

This appears to have been the case in every release I checked (3.5.0 through current main); it's a long-standing omission rather than a recent regression.

Change

• Add "group by" to the keyword tuple in pgcli/packages/sqlcompletion.py so it routes through _suggest_expression, exactly like ORDER BY.
• Extend the existing test_distinct_and_order_by_suggestions_with_aliases and ..._with_alias_given parametrized tests with GROUP BY cases (both bare GROUP BY and qualified GROUP BY x.).
• Add a changelog entry.

Testing

• pytest tests/ → 2699 passed, 1 xfailed, 1 xpassed (pre-existing).
• ruff format --check and ruff check clean on the changed files.