feat(criterion): make Bencher implement Send + Sync

[DaniPopes]

criterion's Bencher implements Send + Sync, which allows running the benchmark itself on another thread. This allows something like |b| /* my rayon thread pool */.install(|| b.iter(|| /* bench */)).

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #116 will degrade performances by 15.89%

_{Comparing DaniPopes:criterion-bencher-send (97babdc) with main (43a2ff0)}

Summary

⚡ 8 improvements
❌ 10 regressions
✅ 149 untouched

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	Iterative	162.5 ns	132.5 ns	+22.64%
⚡	Iterative[20]	162.5 ns	132.5 ns	+22.64%
⚡	Iterative[21]	164.2 ns	134.2 ns	+22.36%
❌	fibonacci_custom_measurement	916.7 ns	945.8 ns	-3.08%
⚡	sleep_100ms	595.3 ns	536.9 ns	+10.86%
⚡	sleep_10ms	566.1 ns	536.9 ns	+5.43%
⚡	sleep_1ms	566.1 ns	536.9 ns	+5.43%
⚡	sleep_50ms	595.3 ns	536.9 ns	+10.86%
❌	from_elem[1024]	1.8 µs	1.9 µs	-3.1%
❌	from_elem_decimal[1024]	1.8 µs	1.9 µs	-3.1%
❌	bench_array1[10]	225 ns	254.2 ns	-11.48%
❌	bench_array1[4]	154.4 ns	183.6 ns	-15.89%
❌	bench_array2[10]	225 ns	254.2 ns	-11.48%
❌	bench_array2[4]	154.4 ns	183.6 ns	-15.89%
❌	init_array[4]	154.4 ns	183.6 ns	-15.89%
❌	recursive[0]	211.1 ns	240.3 ns	-12.14%
❌	recursive[5]	291.1 ns	320.3 ns	-9.11%
⚡	recursive_memoized[BTreeMap<u64, u64>, 0]	430 ns	400.8 ns	+7.28%

[codspeed-hq]

CodSpeed WallTime Performance Report

Merging #116 will degrade performances by 20.4%

_{Comparing DaniPopes:criterion-bencher-send (97babdc) with main (43a2ff0)}

Summary

⚡ 15 improvements
❌ 8 regressions
✅ 131 untouched

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	build_vec	214 ns	222 ns	-3.6%
⚡	large_drop	168.4 µs	161.5 µs	+4.33%
⚡	large_setup	16 ns	12 ns	+33.33%
❌	from_elem[2048]	255 ns	271 ns	-5.9%
⚡	from_elem[4096]	347 ns	334 ns	+3.89%
⚡	generate_combinations[6]	1.4 µs	1.3 µs	+7.28%
⚡	generate_combinations[7]	2.1 µs	2 µs	+5.96%
❌	generate_parentheses[6]	68.3 µs	72.2 µs	-5.5%
⚡	graph_coloring[3]	1.2 µs	1.1 µs	+4.35%
⚡	graph_coloring[4]	962 ns	896 ns	+7.37%
❌	graph_coloring[5]	971 ns	1,025 ns	-5.27%
⚡	hamiltonian_cycle[4]	575 ns	540 ns	+6.48%
❌	hamiltonian_cycle[5]	606 ns	656 ns	-7.62%
⚡	hamiltonian_cycle[7]	899 ns	776 ns	+15.85%
⚡	n_queens_solver[5]	9.1 µs	7.5 µs	+21.96%
❌	permutations[5]	11.6 µs	14.6 µs	-20.4%
⚡	permutations[6]	76.4 µs	68.8 µs	+11.01%
⚡	permutations[7]	792.1 µs	767 µs	+3.27%
❌	rat_in_maze[5]	573 ns	603 ns	-4.98%
❌	rat_in_maze[8]	902 ns	932 ns	-3.22%
...	...	...	...	...

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

[not-matthias]

LGTM!

Minor style nitpick: You could remove the let codspeed = &mut *self.codspeed; variables altogether and use it directly when calling the functions (e.g. self.codspeed.start_benchmark)

[DaniPopes]

Any chance this can get merged and released soon? Thanks! :D @art049 @not-matthias