feat: Support use of ordinal values in SQL ORDER BY clause

[alexander-beedie]

Closes #15793.

Factored out the common code we're using to handle GROUP BY¹ ordinals and used it to integrate support for the same capability in ORDER BY² clauses.

Example

df = pl.DataFrame({
    "a": ["xx", "xx", "yy", "yy"],
    "b": [4000, 3000, 2000, 1000],
})

df.sql("""
  SELECT * 
    FROM self 
    ORDER BY 1 DESC, 2 ASC
""")
# shape: (4, 2)
# ┌─────┬──────┐
# │ a   ┆ b    │
# │ --- ┆ ---  │
# │ str ┆ i64  │
# ╞═════╪══════╡
# │ yy  ┆ 1000 │
# │ yy  ┆ 2000 │
# │ xx  ┆ 3000 │
# │ xx  ┆ 4000 │
# └─────┴──────┘

df.sql("""
  SELECT a, SUM(b) AS sum_b 
    FROM self 
    GROUP BY 1 
    ORDER BY 2 DESC
""")
# shape: (2, 2)
# ┌─────┬───────┐
# │ a   ┆ sum_b │
# │ --- ┆ ---   │
# │ str ┆ i64   │
# ╞═════╪═══════╡
# │ xx  ┆ 7000  │
# │ yy  ┆ 3000  │
# └─────┴───────┘

Footnotes

1. https://github.com/pola-rs/polars/pull/15584 ↩

2. "A sort_expression can also be the column label or number of an output column" https://www.postgresql.org/docs/current/queries-order.html#QUERIES-ORDER ↩

[codspeed-hq]

CodSpeed Performance Report

Merging #16745 will degrade performances by 17.33%

_{Comparing alexander-beedie:sql-ordinal-order-by (25df658) with main (6f3fd8e)}

Summary

❌ 1 regressions
✅ 36 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	alexander-beedie:sql-ordinal-order-by	Change
❌	test_groupby_h2oai_q1	2.1 ms	2.6 ms	-17.33%