test(bench): add concat_source_add microbenchmark

[JSerFeng]

What

Adds two new criterion benchmarks under the existing rspack_sources group:

• concat_source_add_many — builds a ConcatSource by sequentially adding 500 RawStringSource children
• concat_source_add_few — same shape, but with 16 children

Why

The current bench suite builds each ConcatSource once during setup and then loops over .map().to_json() (or .source(), .size(), hashing, etc.) inside b.iter. That measures the cached read path, but leaves the construction path — the repeated add calls — completely uninstrumented.

In real Rspack code generation, add is one of the hottest patterns: rspack_plugin_javascript/src/runtime.rs does a par_iter().fold(ConcatSource::default, |mut acc, src| { acc.add(src); acc }), and rspack_core's concatenated module rendering chains 300+ add calls per module. Without coverage here, perf changes touching add (good or bad) slip past CodSpeed silently — for example, the recent lock() → get_mut() change in add showed no movement on the existing benches even though it's measurably ~6–7 ns faster per call.

The 500/16 sizes were picked to match those two real shapes (concatenated module assembly vs. runtime module wrapping).

How

• benches/bench.rs: two new functions, each building Vec<BoxSource> of N RawStringSource children up-front (so allocation is excluded from the hot loop), then b.iter builds a fresh ConcatSource and adds them all.
• Registered in the existing rspack_sources benchmark group alongside the others.
• No production code changes.

[codspeed-hq]

Merging this PR will not alter performance

✅ 11 untouched benchmarks
🆕 2 new benchmarks
⏩ 7 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
🆕	concat_source_add_few	N/A	6.3 µs	N/A
🆕	concat_source_add_many	N/A	51.2 µs	N/A

---

_{Comparing fy/bench-concat-source-add (678e9a7) with main (7795fd2)}

Footnotes

1. 7 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. ↩

[Copilot]

Pull request overview

Adds Criterion microbenchmarks to measure the ConcatSource construction hot path in rspack_sources, focusing specifically on repeated add calls (a pattern heavily used during Rspack codegen).

Changes:

• Add concat_source_add_many benchmark that builds a fresh ConcatSource by adding 500 children in the timed loop.
• Add concat_source_add_few benchmark that builds a fresh ConcatSource by adding 16 children in the timed loop.
• Register both benchmarks under the existing rspack_sources benchmark group.

---

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