refactor(decimal): dispatch numeric kernels on one storage width

[mhk197]

checked_decimal_arrays is generic over the working width (6 types) and the operator (4), so there are 24 copies of the outer function. Inside each copy, a match_each_decimal_value_type! on the lhs storage type nests another on the rhs storage type, so every copy contains 36 checked_lanes closures. The 24 outer functions plus their 24 × 36= 864 closures make instantiations.

The nesting exists because a DecimalArray's storage width is independent of its precision, so two operands sharing a DecimalDType may still be stored at different widths.

Instead, this PR dispatches on a single storage width. A mismatched pair is first widened to the wider of the two with the existing widened_buffer (zero-copy when they already match), and the lane loop then runs once per (working width, operator, storage width) with the same inline checked cast as before.

Mono items in vortex-array (cargo +nightly rustc -p vortex-array --lib -- -Zprint-mono-items):

	before	after
checked_decimal_arrays	888	168
numeric::decimal::* module total	1,372	652
lane_kernels::* total	15,544	9,694
whole crate	100,083	93,513

Matched-storage inputs take the code path they did before, so the binary_ops decimal benchmarks are unchanged within noise.

[codspeed-hq]

Merging this PR will regress 1 benchmark

⚠️ 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.

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 5 improved benchmarks
❌ 1 regressed benchmark
✅ 2175 untouched benchmarks
⏩ 206 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	random_i8[0.5]	67.2 µs	91.3 µs	-26.37%
⚡	Simulation	random_i16[0.8]	92.1 µs	74.4 µs	+23.81%
⚡	Simulation	compact_sliced[(1024, 90)]	1.5 µs	1.3 µs	+16.67%
⚡	Simulation	compact_sliced[(2048, 90)]	1.6 µs	1.4 µs	+15.95%
⚡	WallTime	mul_u32_nonnull_avx512	6.3 µs	5.6 µs	+11.72%
⚡	Simulation	allocate_drop_bytes[0]	520.2 ns	466 ns	+11.62%

Tip

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

---

_{Comparing mk/decimal-kernel-monomorphization (8927d20) with develop (dab1684)}

Footnotes

1. 206 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. ↩