Use loop form in orderedMin4 to enable inlining

[aalpar]

Summary

• Rewrites orderedMin4 from unrolled-with-early-exits to a loop, reducing the Go compiler's inlining cost from 130 to 63 (budget: 80)
• This allows orderedMin4 to be inlined into orderedBubbledown's hot loop, eliminating function call overhead on every iteration
• No algorithm or semantic changes — same min-max heap behavior

Benchmarks

Isolated change, Apple M4 Max, go test -bench BenchmarkOrdered -benchmem -count=5 -benchtime=2s:

                  │   master     │     orderedMin4 loop                │
                  │    sec/op    │    sec/op     vs base               │
OrderedPop-16       286.0n         236.5n       -17.31% (p=0.008)
OrderedPopMax-16    283.9n         234.9n       -17.26% (p=0.008)
OrderedPushPop-16   290.4n         246.5n       -15.12% (p=0.008)
OrderedPush-16      11.83n         11.76n            ~ (p=0.381)

Two other candidate optimizations (isMinHeap simplification, bounds-check-elimination hint in orderedBubbledown) were benchmarked in isolation and showed slight regressions, so they were dropped.

Test plan

• go test -race -count=1 ./... passes
• Existing unit, randomized (1000 iterations), and fuzz tests all exercise orderedMin4 through Pop/PopMax/Remove/Fix paths

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR optimizes the generic min/max heap’s hot path by rewriting orderedMin4 into a loop form that stays within the Go compiler’s inlining budget, enabling inlining into orderedBubbledown and reducing per-iteration overhead.

Changes:

• Replaces the unrolled/early-exit implementation of orderedMin4 with a bounded loop over up to 4 consecutive elements.
• Adds documentation explaining the inlining-cost motivation for the loop form.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.