Split up binary ScalarFn

[gatesn]

Edit: im not convinced of this PR, but the benchmarks have surfaced some interesting results. So digging in

---

Adds new scalar functions for comparison, arithmetic, and logical operators.

Previously the single binary function typically just switched on the logic for each of these, making it much harder to read and follow.

This also gives us more granular kernel selection, e.g. for arrays that wish to provide a comparison kernel we would previously have to invoke the kernel for any binary function.

[gatesn]

I hate this. we should just rewrite into the new kernel traits

[codspeed-hq]

Merging this PR will degrade performance by 92.63%

⚡ 6 improved benchmarks
❌ 10 regressed benchmarks
✅ 993 untouched benchmarks
⏩ 1515 skipped benchmarks¹

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

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	pushdown_compare[(1000, 64, 4)]	473.7 µs	390.6 µs	+21.26%
⚡	Simulation	pushdown_compare[(1000, 4, 4)]	472.8 µs	217.2 µs	×2.2
⚡	Simulation	pushdown_compare[(1000, 16, 8)]	495.4 µs	253.4 µs	+95.55%
⚡	Simulation	pushdown_compare[(1000, 16, 4)]	473.3 µs	243.2 µs	+94.63%
❌	Simulation	pushdown_compare[(10000, 16, 8)]	682.4 µs	1,548.1 µs	-55.92%
⚡	Simulation	pushdown_compare[(1000, 4, 8)]	474.5 µs	220.6 µs	×2.2
❌	Simulation	pushdown_compare[(10000, 16, 4)]	634 µs	1,438.4 µs	-55.92%
❌	Simulation	pushdown_compare[(10000, 4, 4)]	634 µs	1,170.5 µs	-45.84%
⚡	Simulation	pushdown_compare[(1000, 64, 8)]	484.4 µs	434 µs	+11.63%
❌	Simulation	pushdown_compare[(10000, 64, 8)]	644.5 µs	3,353.9 µs	-80.78%
❌	Simulation	pushdown_compare[(10000, 4, 8)]	643.2 µs	1,190.8 µs	-45.98%
❌	Simulation	pushdown_compare[(10000, 64, 4)]	633.6 µs	2,913.4 µs	-78.25%
❌	Simulation	old_alp_prim_test_between[f64, 16384]	332.9 µs	372.3 µs	-10.59%
❌	Simulation	old_alp_prim_test_between[f64, 32768]	541.4 µs	609.8 µs	-11.2%
❌	Simulation	eq_pushdown_low_match	2.3 ms	30.9 ms	-92.63%
❌	Simulation	eq_pushdown_high_match	2.4 ms	31.3 ms	-92.34%

---

_{Comparing ngates/split-binary-fn (8d922f4) with develop (fc4d111)}

Footnotes

1. 1515 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. ↩