[codex] Lazily create route transform worker slots

[ScriptedAlchemy]

Summary

• Lazily allocate route transform worker slots instead of creating the full pool the first time the executor is used.
• Keep task scheduling stable while only creating worker instances for slots that receive scheduled work.
• Add focused test coverage for lazy worker-slot creation.

Why

The original worker executor created every route transform worker at the first parallel transform touchpoint. Lazy slot allocation reduces the dev startup cliff and gives the follow-up prewarm PR a cheaper baseline to build from.

Validation

Validated before stacking this PR:

• pnpm exec tsc --noEmit
• pnpm test tests/parallel-route-transforms.test.ts tests/index.test.ts tests/features.test.ts tests/build-output-transforms.test.ts
• pnpm build

[github-actions]

Benchmark Results

Compared PR head 00d0e65 against base 423843e.

Dev Rollup

Group	Fixtures	Base total	Head total	Delta	Base ready	Head ready	Ready delta	Base routes	Head routes	Route delta	Base update/HMR	Head update/HMR	Update delta	Speedup
All dev fixtures	7	33.19s	33.18s	-0.0%	21.92s	21.94s	+0.1%	3.94s	3.93s	-0.4%	3.44s	3.42s	-0.5%	1.00x
Large app	1	14.84s	14.87s	+0.2%	8.88s	8.91s	+0.3%	2.03s	2.02s	-0.5%	2.02s	2.01s	-0.7%	1.00x
Standard fixtures	6	18.35s	18.31s	-0.2%	13.04s	13.03s	-0.1%	1.91s	1.91s	-0.2%	1.42s	1.41s	-0.3%	1.00x

Production Build Benchmarks

Rendered 7 production build benchmarks.

Benchmark	Runs	Base total	Head total	Delta	Head mean	Head p95	Speedup	Head RSS p95
large-355-ssr-esm	5	9.49s	9.49s	-0.0%	9.54s	9.82s	1.00x	1567 MB
synthetic-1024-ssr-esm	5	4.96s	4.81s	-3.1%	4.84s	5.07s	1.03x	625 MB
synthetic-1024-ssr-esm-split	5	6.23s	6.31s	+1.4%	6.32s	6.47s	0.99x	794 MB
synthetic-256-sourcemaps	10	2.90s	2.91s	+0.5%	2.92s	3.10s	1.00x	482 MB
synthetic-256-ssr-esm	10	2.86s	2.77s	-3.0%	2.79s	3.04s	1.03x	442 MB
synthetic-256-ssr-esm-split	10	3.20s	3.23s	+1.0%	3.23s	3.42s	0.99x	489 MB
synthetic-48-ssr-esm	10	1.86s	1.83s	-1.3%	1.87s	2.22s	1.01x	338 MB

full Dev Fixture Summary

Rendered 7 dev benchmark fixtures from the full profile.

Benchmark	Runs	Base total	Head total	Delta	Base ready	Head ready	Base routes	Head routes	Base update/HMR	Head update/HMR	Update delta	Head mean	Head p95	Speedup	Head RSS p95
large-355-ssr-esm	5	14.84s	14.87s	+0.2%	8.88s	8.91s	2.03s	2.02s	2.02s	2.01s	-0.7%	14.88s	15.10s	1.00x	-
synthetic-1024-ssr-esm	5	5.07s	5.10s	+0.5%	3.41s	3.44s	0.54s	0.54s	0.48s	0.48s	-0.3%	5.11s	5.22s	0.99x	-
synthetic-1024-ssr-esm-split	5	5.13s	5.07s	-1.3%	3.46s	3.41s	0.54s	0.54s	0.48s	0.48s	-0.5%	5.12s	5.25s	1.01x	-
synthetic-256-sourcemaps	10	2.51s	2.47s	-1.6%	1.91s	1.89s	0.23s	0.23s	0.15s	0.15s	-0.1%	2.48s	2.55s	1.02x	-
synthetic-256-ssr-esm	10	2.26s	2.28s	+0.9%	1.71s	1.72s	0.23s	0.23s	0.13s	0.13s	-0.1%	2.28s	2.35s	0.99x	-
synthetic-256-ssr-esm-split	10	2.26s	2.28s	+0.8%	1.69s	1.72s	0.23s	0.23s	0.13s	0.13s	+0.2%	2.30s	2.38s	0.99x	-
synthetic-48-ssr-esm	10	1.10s	1.11s	+0.5%	0.85s	0.84s	0.13s	0.13s	0.05s	0.05s	-0.9%	1.10s	1.12s	1.00x	-

Synthetic Rsbuild App

Rendered 2 production build benchmarks.

Benchmark	Runs	Base total	Head total	Delta	Head mean	Head p95	Speedup	Head RSS p95
complex app	2	98.03s	104.10s	+6.2%	104.10s	-	0.94x	-
complex app	2	102.13s	104.46s	+2.3%	104.46s	-	0.98x	-

Rendered 1 dev benchmark fixture from the embedded complex app.

Benchmark	Runs	Base total	Head total	Delta	Base ready	Head ready	Base routes	Head routes	Base update/HMR	Head update/HMR	Update delta	Head mean	Head p95	Speedup	Head RSS p95
complex app	2	89.51s	90.55s	+1.2%	79.61s	80.40s	3.09s	3.27s	4.33s	4.33s	-0.0%	90.55s	-	0.99x	-

Profile: full; mode: dev; iterations: 10; warmup: 0.
Workflow run