Add array subset-check benchmark across Ruby 3.3, 3.4, 4.0

[etagwerker]

What

Adds an Array benchmark for subset checks (is every element of a1 also in a2?), comparing five approaches across Ruby 3.3.10, 3.4.7, and 4.0.0:

• (a1 - a2).empty?
• (a1 & a2) == a1
• (a1 & a2).size == a1.size
• a1.all? { |e| a2.include?(e) }
• a1.to_set.subset?(a2.to_set)

Background

This revisits the comparison from #125 by @gabteles (now closed). That PR had two problems the reviewers (@mblumtritt, @Arcovion) hinted at back in 2017:

1. A correctness bug: Set#subset? had its arguments reversed (a2.to_set.subset?(a1.to_set)), so it returned false while every other method returned true. It wasn't measuring the same operation.
2. No stable winner: the result is highly data-dependent.

This version fixes the Set arguments, adds an equivalence guard so all five approaches must agree before the benchmark runs, and reports results across three modern Ruby versions.

Findings

• (a1 - a2).empty? is the consistent winner across 3.3, 3.4, and 4.0 for the common case where a1 really is a subset.
• a1.all? { include? } is data-dependent: it short-circuits on the first miss (so it wins when a1 is not a subset), but it's O(n*m) and degrades badly on large true subsets. The README entry documents this caveat.
• Set#subset? improved dramatically in Ruby 4.0: ~6.8x slower on 3.3/3.4 (dominated by to_set allocation) but only ~1.7x slower on 4.0. If you already hold Sets or check repeatedly, it scales best.

Notes

• Benchmarks were run via rbenv on each version (benchmark-ips installed per version).
• The README block shows the full output for 3.3.10 and the Comparison: summary for 3.4.7 and 4.0.0 to keep it readable.

🤖 Generated with Claude Code