Add priority heap benchmarks

[samuel-williams-shopify]

Summary

• Add benchmark/io/event/priority_heap.rb using sus-fixtures-benchmark.
• Cover priority heap push, concat, pop, heapify, and delete_if workloads.
• Capture the findings from evaluating a targeted native extension prototype against Ruby and YJIT.

Findings

A targeted native prototype patched the hot PriorityHeap methods from the extension. It improved plain Ruby in several heap-heavy paths, but YJIT generally performed better without adding C maintenance or load-order complexity. The native prototype is not included in this PR; this PR keeps the benchmark so future changes can be evaluated consistently.

Local benchmark results from benchmark/io/event/priority_heap.rb:

Benchmark	Ruby	Ruby + YJIT	Native Prototype
push in-order	573.35ns	469.49ns	512.48ns
push reverse-order	1.19us	752.48ns	695.09ns
concat into empty heap	467.75us	81.11us	131.37us
pop	1.97us	757.72ns	861.19ns
heapify after delete/append	1.17ms	335.63us	485.32us
delete_if half	1.05ms	383.57us	420.06us

Conclusion: YJIT is the better optimization path for the current Ruby heap implementation. A native implementation does not look like a net win unless non-YJIT performance becomes important enough to justify maintaining duplicate heap logic in C.

Verification

• BUNDLE_WITH='' bundle exec rubocop benchmark/io/event/priority_heap.rb
• BUNDLE_WITH='' bundle exec sus --verbose benchmark/io/event/priority_heap.rb
• BUNDLE_WITH='' bundle exec ruby --yjit -S sus --verbose benchmark/io/event/priority_heap.rb