perf: resolve RSpec test file paths without allocating per example

[connorshea]

Description

AI Disclosure: This was generated using Claude Code with Opus 5. It has been reviewed and tested by me manually.

RSpecAdapter.file_path_for runs on every test example (in Queue Mode, it runs twice per example) and was allocating more objects than necessary to check candidate paths. We can instead check it inline while still keeping the laziness.

We built a MatchData and a few other things for every call of parse_file_path and id_path?. We don't need to do that here, because we can instead look for the last [ in the path and then match on the string from there onward, without needing to allocate almost anything.

parse_file_path now finds [ with rindex and validates it with an allocation-free Regexp#match? starting at that offset, and id_path? is now an end_with? + match?.

The only functional change in this code is that a path with a linebreak (which shouldn't happen anyway) would previously cause a NoMethodError, and is now ignored/kept instead. I don't think that's a problem because it doesn't ever happen, but figured I'd note it.

Also wanted to note that I had Claude fuzz the new regex to ensure equivalent behavior on 250k values (some generated to match real paths, some fully random values).

---

This saves a tiny bit of time, but the main benefit of this change is reducing the # of object allocations because it helps avoid triggering GC in large suites, which speeds things up. This isn't going to be hugely significant on its own, but I think it's worth doing.

Per-call, Ruby 3.4.9, arm64-darwin25, 500 RSpec example ids, benchmark-ips (3s, 1s warmup)

Method	Objects before	Objects after	µs before	µs after	Speedup
file_path_for	13	1	0.926	0.206	4.5x
parse_file_path (id path)	6	1	0.531	0.153	3.5x
parse_file_path (plain path)	5	0	0.493	0.040	12.4x
id_path? (id path)	6	0	0.541	0.089	6.1x
id_path? (plain path)	5	0	0.511	0.038	13.5x

Scaled to a suite (one file_path_for per example, NOT Queue Mode)

Examples	Time before	Time after	Objects before	Objects after
10,000	9.1 ms	2.0 ms	130,007	10,007
100,000	94.6 ms	20.6 ms	1,300,000	100,000

Scaled to a suite (Queue Mode: two file_path_for calls per example)

Examples	Time before	Time after	Objects before	Objects after
10,000	18.2 ms	4.0 ms	260,014	20,014
100,000	189.2 ms	41.2 ms	2,600,000	200,000

Checks

• I added the changes to the UNRELEASED section of the CHANGELOG.md, including the needed bump (i.e., patch, minor, major)
• I followed the architecture outlined below for RSpec in Queue Mode:
  • Pure: lib/knapsack_pro/pure/queue/rspec_pure.rb contains pure functions that are unit tested.
  • Extension: lib/knapsack_pro/extensions/rspec_extension.rb encapsulates calls to RSpec internals and is integration and E2E tested.
  • Runner: lib/knapsack_pro/runners/queue/rspec_runner.rb invokes the pure code and the extension to produce side effects, which are integration and E2E tested.