Speed up scope parsing, especially for deep scopes

[sirosen]

There are three distinct optimizations here, applied in series:

• Make scope parsing faster on deep scope strings
• Remove unused scope graph 'nodes' set
• Prefer 'reversed' over reversing slice

By way of explanation, the first optimization, which does some graph-coloring with an enum to search for back-edges.
This replaces the prior implementation, which made sets of "all ancestors" for each node, which results in a large number of new sets being built, especially for deeper scopes.

The micro-optimizations are very minor cleanup (the commits should be self-explanatory).

[codspeed-hq]

Merging this PR will improve performance by ×3.9

⚡ 5 improved benchmarks
✅ 13 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_deep_scope_parsing[5000]	2,570.1 ms	398.8 ms	×6.4
⚡	test_deep_scope_parsing[4000]	1,606.2 ms	317.5 ms	×5.1
⚡	test_deep_scope_parsing[3000]	1,019.4 ms	236.9 ms	×4.3
⚡	test_deep_scope_parsing[2000]	478.2 ms	156.6 ms	×3.1
⚡	test_deep_scope_parsing[1000]	155.2 ms	76.5 ms	×2

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing sirosen:faster-scope-parsing (609c52b) with main (d32735a)}