[FEAT] [SQL] Add global agg support for SQL

[amitschang]

This adds support to SQL to run global aggregation when there are only aggregations and no group by, which is inline with many other SQL implementations.

It is similar in nature to #1979, but specific to SQL - depending on how that issue is resolved the SQL planner code here could be redundant.

Example:

d = daft.from_pydict({"i": [1,2,3,4,5]})
daft.sql('''SELECT sum(i) FROM d''', SQLCatalog({'d': d})).collect()
╭───────╮                                                                                                    
│ i     │
│ ---   │
│ Int64 │
╞═══════╡
│ 15    │
╰───────╯

note: I know the SQL support is new, so sorry if this is jumping the gun! Is it better to first create discussions or issues for this kind of change? I've been looking forward to SQL in this project and from playing around with using sqlglot on the python side to build a logical plan, I noticed a few things - now that this is here I thought I might contribute 😄

[amitschang]

I'm not totally sure this gets run in make test (and running via rust-analyzer didn't work out-of-the-box)

[codspeed-hq]

CodSpeed Performance Report

Merging #2799 will degrade performances by 18.98%

_{Comparing amitschang:sql-global-agg (d2230f9) with main (91d9fe9)}

Summary

⚡ 2 improvements
❌ 1 regressions
✅ 13 untouched benchmarks

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

Benchmarks breakdown

	Benchmark	main	amitschang:sql-global-agg	Change
❌	test_count[1 Small File]	19.6 ms	24.2 ms	-18.98%
⚡	test_explain[100 Small Files]	49 ms	37.6 ms	+30.14%
⚡	test_show[100 Small Files]	307.8 ms	49.7 ms	×6.2

[djouallah]

I would really appreciate a syntax like this just for compatibility reason with duckdb/datafusion/polars/chdb

@jaychia let's just make life easy for new users

something like this

con = SQLCatalog({"test": daft.from_pydict({"a": [1, 2, 3]})})
con .sql("SELECT * FROM test").show()

[universalmind303]

I would really appreciate a syntax like this just for compatibility reason with duckdb/datafusion/polars/chdb

@jaychia let's just make life easy for new users

something like this

con = SQLCatalog({"test": daft.from_pydict({"a": [1, 2, 3]})})
con .sql("SELECT * FROM test").show()

@djouallah we do have an open issue for selecting from python objects (similar to duckdb and polars) that is pretty high on our priority list!