[red-knot] Method calls and the descriptor protocol + 16268

[sharkdp]

Summary

#16121 + #16268

Test Plan

[codspeed-hq]

CodSpeed Performance Report

Merging #16279 will degrade performances by 24.12%

_{Comparing david/descriptor-protocol-merge-16268 (9b93739) with main (b385c7d)}

Summary

❌ 2 regressions
✅ 30 untouched benchmarks

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	red_knot_check_file[cold]	91.7 ms	101.3 ms	-9.51%
❌	red_knot_check_file[incremental]	5.7 ms	7.6 ms	-24.12%

[sharkdp]

The first attempt is worse than my PR alone:

	Benchmark	BASE	HEAD	Change
❌	red_knot_check_file[cold]	91.7 ms	104.6 ms	-12.39%
❌	red_knot_check_file[incremental]	5.7 ms	8.2 ms	-29.93%

[sharkdp]

It looks better when symbol_by_id is a query again:

	Benchmark	BASE	HEAD	Change
❌	red_knot_check_file[cold]	91.7 ms	101.3 ms	-9.5%
❌	red_knot_check_file[incremental]	5.7 ms	7.6 ms	-24.12%

[MichaReiser]

That's good to know. It shows that symbol_by_id being a query is important

[MichaReiser]

It now makes me wonder if #16265 would perform even better?

[sharkdp]

It now makes me wonder if #16265 would perform even better?

No: #16280 (comment)