Run code generation on build

[Wesley Wigham (weswigham)]

So that you never need to remember "oh, I changed X so need to regenerate Y" ever again.

Also fixes build:watch to do code generation, too, if people are using that.

[Copilot]

Copilot review overview

🟡 Changes recommended

Watch reruns can bypass schema validations and mark stale generator output as current.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 2

Open (2)

• Reload changed generator implementations before watch regeneration · New
• Rerun NodeBase validation checks after each schema reload · New

What changed in this PR

Runs code generation automatically before builds and during watch mode.

Changes:

• Adds generated-output metadata and schema reloading.
• Integrates generation into build, API, validation, watch, and release tasks.
• Adds watch-mode coverage and updates contributor guidance.

File	Description
Herebyfile.mjs	Integrates generation with build and watch tasks.
CONTRIBUTING.md	Documents the revised workflow.
tools/​scripts/​gen/​generatedFile.test.mts	Tests repeated watch generation.
tools/​scripts/​tsc/​schema.ts	Supports schema reloading.
tools/​scripts/​tsc/​generate.ts	Reloads schema before generation.
tools/​scripts/​tsc/​generate-ts-ast.ts	Refreshes schema-derived state and exports outputs.
tools/​scripts/​tsc/​generate-go-ast.ts	Exports generated output paths.
tools/​scripts/​tsc/​generate-enums.ts	Exports enum output paths.
tools/​scripts/​tsc/​generate-encoder.ts	Exports encoder output paths.
tsc/​internal/​stringutil/​_scripts/​generate-unicode-data.mts	Exports Unicode output paths.

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[Wesley Wigham (weswigham)]

It's a bit of a tangent, but I also just went ahead and deleted all the tests in generatedFile.test.mts that were more like npx hereby task integration tests than caching functionality tests - while I get where they came from, if they have a place (though I think they do not), it's not there. They were getting progressively weirder in how specifically they were trying to introspect and lock-down task functionality with every new test, but I fundamentally don't think there's a good place to "test" top-level task definition functionality like that.