Add execution target that recursively canonicalises as well as canonical with canonical validity

[robert3005]

Should there be an explicit option to execute validity just by one step?

[codspeed-hq]

Merging this PR will degrade performance by 99.99%

⚡ 1 improved benchmark
❌ 52 regressed benchmarks
✅ 1082 untouched benchmarks
⏩ 1268 skipped benchmarks¹

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

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	take_fsl_random[1024, 100]	6.6 µs	2,002.5 µs	-99.67%
❌	Simulation	take_fsl_nullable_random[256, 1000]	6.6 µs	4,739.6 µs	-99.86%
❌	Simulation	take_fsl_nullable_random[64, 100]	6.6 µs	178 µs	-96.28%
❌	Simulation	take_fsl_nullable_random[1024, 1000]	6.6 µs	22,598.3 µs	-99.97%
❌	Simulation	take_fsl_nullable_random[1024, 100]	6.6 µs	1,999.7 µs	-99.67%
❌	Simulation	take_fsl_nullable_random[16, 1000]	6.6 µs	411.7 µs	-98.41%
❌	Simulation	take_fsl_random[1024, 1000]	6.6 µs	23,032.3 µs	-99.97%
❌	Simulation	take_fsl_random[16, 100]	6.6 µs	70.6 µs	-90.63%
❌	Simulation	take_fsl_random[16, 1000]	6.6 µs	377.4 µs	-98.26%
❌	Simulation	take_fsl_nullable_random[16, 100]	8 µs	96.2 µs	-91.71%
❌	Simulation	take_fsl_nullable_random[4096, 100]	6.6 µs	8,104.3 µs	-99.92%
❌	Simulation	take_fsl_nullable_random[256, 100]	6.6 µs	538.1 µs	-98.78%
❌	Simulation	take_fsl_random[4096, 100]	6.6 µs	8,191.5 µs	-99.92%
❌	Simulation	take_fsl_nullable_random[4096, 1000]	6.6 µs	90,781.9 µs	-99.99%
❌	Simulation	take_fsl_random[256, 100]	6.6 µs	530.6 µs	-98.75%
❌	Simulation	take_fsl_nullable_random[64, 1000]	6.6 µs	1,277.2 µs	-99.48%
❌	Simulation	take_fsl_random[256, 1000]	6.6 µs	4,933.7 µs	-99.87%
❌	Simulation	take_fsl_random[4096, 1000]	6.6 µs	92,697.2 µs	-99.99%
❌	Simulation	take_fsl_random[64, 100]	6.6 µs	162.4 µs	-95.95%
❌	Simulation	take_fsl_random[64, 1000]	6.6 µs	1,288.3 µs	-99.49%
...	...	...	...	...	...

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

---

_{Comparing rk/recursivecanonical (90b387b) with develop (6135037)}

Footnotes

1. 1268 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]

would be better if the ctx was created outside of the bench

[joseph-isaacs]

cna we do this without an alloc?

[robert3005]

sadly no, we could store fields though as Arc<Box<[ArrayRef]>>. Otherwise this only works with unsized locals since after unwrap you get [ArrayRef] and compiler doesn't know what to do with it

[joseph-isaacs]

Looks good, just one Q