perf: remove unneeded scalar_at and valid_at and use arrays instead

[joseph-isaacs]

We often pull out and decompress array when doing is_valid in a loop, this PR instead uses execute_mask once and then operates on a Mask in the loop.

We could later come along and replace this Mask::value(&self) method with a packed chunk iterator.

[codspeed-hq]

Merging this PR will improve performance by 20.38%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 6 improved benchmarks
❌ 2 regressed benchmarks
✅ 1511 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	chunked_varbinview_opt_canonical_into[(1000, 10)]	191.4 µs	226.9 µs	-15.65%
❌	Simulation	chunked_varbinview_opt_into_canonical[(1000, 10)]	205.7 µs	241.5 µs	-14.83%
⚡	Simulation	extend_from_array_non_zctl_overlapping[(10000, 8)]	4 ms	2.3 ms	+70.87%
⚡	Simulation	extend_from_array_non_zctl_overlapping[(1000, 8)]	454 µs	286.7 µs	+58.35%
⚡	Simulation	chunked_bool_canonical_into[(1000, 10)]	46.9 µs	31.7 µs	+48.13%
⚡	Simulation	extend_from_array_non_zctl_overlapping[(1000, 32)]	949.8 µs	781.9 µs	+21.48%
⚡	Simulation	chunked_varbinview_canonical_into[(100, 100)]	311.3 µs	274.5 µs	+13.42%
⚡	Simulation	chunked_varbinview_into_canonical[(100, 100)]	363.2 µs	326.7 µs	+11.17%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing ji/row-accessor-remove (b555fd5) with develop (998bc65)}