Run fixpoint per strongly connected component

[MichaReiser]

Run a single fixpoint iteration for strongly connected components
to avoid exponential runtime for deeply nested cycles.

TODO:

• Documentation
• Investigate perf regressions
• Test that demonstrates the "collapsing" of nested cycles
• Review if there are any changes necessary to maybe_changed_after
• Investigate why cycle_nested_deep_conditional_changed::the_test sometimes hangs or panics

[netlify]

✅ Deploy Preview for salsa-rs canceled.

Name	Link
🔨 Latest commit	15e650a
🔍 Latest deploy log	https://app.netlify.com/projects/salsa-rs/deploys/68dcd71bb0779f0008ed6c0b

[codspeed-hq]

CodSpeed Performance Report

Merging #995 will degrade performances by 11.27%

_{Comparing MichaReiser:fixpoint-scc (d7e0774) with master (9c3278f)}

Summary

❌ 4 regressions
✅ 8 untouched

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	amortized[Input]	2.3 µs	2.4 µs	-4.18%
❌	amortized[InternedInput]	2.2 µs	2.3 µs	-4.34%
❌	amortized[SupertypeInput]	2.9 µs	3.1 µs	-6.35%
❌	converge_diverge	127.3 µs	143.4 µs	-11.27%

[MichaReiser]

Assuming I'm getting this to work. One interesting optimization for memo.block_on is to have a DependencyGraph::all_cycles(queries) method instead. That's much cheaper than iterating over all memos (and only requires aquiring a single lock)