perf: avoid type checks in memo and field lookups

[MichaReiser]

Memo and field lookups repeated TypeId checks even though the database-bound ingredient fixes the accessed type. Bind each ingredient to its database context and acquire thread-local state only for operations that need it.

Fixed hot memo hits drop from 233 to 228 instructions (-2.15%), untracked tracked-struct field bodies drop from 59 to 58, CodSpeed reports a 6.76% aggregate improvement, and ty's 29-case suite improves by 0.145% weighted overall.

Testing: Full target, manual-registration, lint, and performance suites passed.

[netlify]

✅ Deploy Preview for salsa-rs canceled.

Name	Link
🔨 Latest commit	7045e87
🔍 Latest deploy log	https://app.netlify.com/projects/salsa-rs/deploys/6a4b72ff763c39000814fb7d

[codspeed-hq]

Merging this PR will improve performance by ×3.3

⚡ 8 improved benchmarks
❌ 5 regressed benchmarks
✅ 61 untouched benchmarks
⏩ 12 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	cold_execute	1.6 ms	1.9 ms	-14.72%
❌	WallTime	parallel_fast_path[NoEviction]	365.4 µs	423.4 µs	-13.7%
❌	WallTime	project_check_then_incremental[NoEviction]	6 ms	6.9 ms	-11.8%
❌	WallTime	parallel_fast_path[Lru]	3.7 ms	4.1 ms	-9.33%
❌	WallTime	project_check_then_incremental[Lru]	9.9 ms	10.8 ms	-8.22%
⚡	Simulation	intern_distinct_many	24,550.2 µs	128.3 µs	×190
⚡	Simulation	tracked_reuse_few	3,604.1 µs	102.2 µs	×35
⚡	Simulation	tracked_reuse_many	986.3 µs	83.8 µs	×12
⚡	Simulation	untracked_field_read_many	1,281.9 µs	217.7 µs	×5.9
⚡	Simulation	tracked_field_read_many	1,260.7 µs	266.6 µs	×4.7
⚡	Simulation	input_field_read_many	559.8 µs	257.6 µs	×2.2
⚡	Simulation	interned_field_read_many	324.9 µs	211.1 µs	+53.91%
⚡	Simulation	fill_and_evict[Lru]	548.8 µs	442.1 µs	+24.16%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing MichaReiser:optimize-memo-lookup (1914bcd) with master (af2381a)}

Footnotes

1. 12 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[MichaReiser]

This is cool but I don't think codex's reasoning here is safe. The issue is that zalsa could be smuggled in. So we'd have to make fetch unsafe.