perf[buffer]: iteration for fallible operations with validity

[joseph-isaacs]

Currently use (and arrow) handle fallible operations with scalar (non-SIMD) code.

This PR add a trait and methods to have fast SIMD checked operations (includes cast) but verified else where that checked_add benefits

[codspeed-hq]

Merging this PR will not alter performance

✅ 1227 untouched benchmarks
🆕 10 new benchmarks
⏩ 39 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
🆕	Simulation	cast_u32_to_u8[65536]	N/A	250.5 µs	N/A
🆕	Simulation	cast_i32_to_u32[65536]	N/A	832.9 µs	N/A
🆕	Simulation	cast_u16_to_u32[65536]	N/A	210.6 µs	N/A
🆕	Simulation	try_map_masked_into_in_place_narrow_i32_u32[65536]	N/A	172.7 µs	N/A
🆕	Simulation	bitpack_value_only[65536]	N/A	452 µs	N/A
🆕	Simulation	map_with_mask_widen_u16_u32[65536]	N/A	189.6 µs	N/A
🆕	Simulation	map_with_mask_narrow_u64_u32[65536]	N/A	387.1 µs	N/A
🆕	Simulation	try_map_into_narrow_u64_u32[65536]	N/A	424 µs	N/A
🆕	Simulation	try_map_masked_into_narrow_i32_u32[65536]	N/A	292.4 µs	N/A
🆕	Simulation	try_map_masked_into_widen_u16_u32[65536]	N/A	190 µs	N/A

---

_{Comparing ji/fast-iter-valid (ca2ad88) with develop (a2323f1)}

Footnotes

1. 39 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[joseph-isaacs]

Open question is where to put this code?