chore: Move rolling_sum cast for small integer types to logical plan #21421

Performance Gate Passed

🆕 41 new benchmarks

	Benchmark	`BASE`	`HEAD`	Change
🆕	`test_to_numpy_series_chunked`	N/A	270.7 µs	N/A
🆕	`test_to_numpy_series_with_nulls`	N/A	343.8 µs	N/A
🆕	`test_to_numpy_series_zero_copy`	N/A	129.5 µs	N/A
🆕	`test_filter1`	N/A	734.4 µs	N/A
🆕	`test_groupby_h2oai_q1`	N/A	1.9 ms	N/A
🆕	`test_groupby_h2oai_q10`	N/A	5.6 ms	N/A
🆕	`test_groupby_h2oai_q2`	N/A	3.8 ms	N/A
🆕	`test_groupby_h2oai_q3`	N/A	2 ms	N/A
🆕	`test_groupby_h2oai_q4`	N/A	1.8 ms	N/A
🆕	`test_groupby_h2oai_q5`	N/A	1.8 ms	N/A
🆕	`test_groupby_h2oai_q6`	N/A	9 ms	N/A
🆕	`test_groupby_h2oai_q7`	N/A	1.9 ms	N/A
🆕	`test_groupby_h2oai_q8`	N/A	2.5 ms	N/A
🆕	`test_groupby_h2oai_q9`	N/A	20.1 ms	N/A
🆕	`test_write_read_scan_large_csv`	N/A	37.8 ms	N/A
🆕	`test_non_strict_inequalities`	N/A	189.6 ms	N/A
🆕	`test_single_inequality`	N/A	77.2 ms	N/A
🆕	`test_strict_inequalities`	N/A	183.9 ms	N/A
🆕	`test_pdsh_q1`	N/A	24.7 ms	N/A
🆕	`test_pdsh_q10`	N/A	9.4 ms	N/A
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.