Monitor only the resolve step in the sudoku bench

[x-hgg-x]

This PR modifies the sudoku bench to only monitor the resolve step.

[codspeed-hq]

CodSpeed Performance Report

Merging #285 will improve performances by ×4

_{Comparing x-hgg-x:sudoku-bench (c319a56) with dev (4c6cdde)}

Summary

⚡ 2 improvements
✅ 4 untouched benchmarks

Benchmarks breakdown

	Benchmark	dev	x-hgg-x:sudoku-bench	Change
⚡	sudoku-easy	15.3 ms	3.8 ms	×4
⚡	sudoku-hard	15.5 ms	3.9 ms	×4

[Eh2406]

This seems reasonable, @konstin up to you.

[konstin]

Some more general thoughts: For real benchmarks, we should track both the setup time and the runtime, since the user has to pay both. But since the sudoku is more a gadget (we can take it out of codspeed now that we have better benchmarks), it's fine to only track the resolve step here.