Format generated vocab file before rename to avoid race condition

[moreal]

Summary

packages/testing test:bun: ../vocab build:self: error: The module's source code could not be parsed: Expected ',', got '<eof>' at file:///home/runner/work/fedify/fedify/packages/vocab/src/vocab.ts:62464:38
packages/testing test:bun: ../vocab build:self:             "The followedMessage must b

According to logs that I see, it failed to run tsdown or cargo check by malformed vocab.ts. And it occurs when 2 or more tasks run the build:self script of @fedify/vocab package. It looks like race condition. And there is only one write behavior to vocab.ts directly. It is deno fmt.

So this pull request makes packages/vocab/scripts/codegen.ts script generate vocab.ts atomically. The atomic means there will be no write behavior on vocab.ts directly. It'll format vocab-{uuid}.ts with deno fmt ... and rename it to vocab.ts.

Related issue

I couldn't find related issue when searching with:

• label:component/ci
• "race"
• "codegen"

Changes

• Let packages/vocab/scripts/codegen.ts script generate src/vocab.ts which already formatted by deno fmt <generatedPath>
• Let deno task compile in @fedify/vocab do not deno fmt. (moved to scripts/codegen.ts script)

Benefits

I believe this change will make CI more consistently from failing by codegen-related race condition.

Checklist

• Did you add a changelog entry to the CHANGES.md?
  • It doesn't change or implement public features.
• Did you write some relevant docs about this change (if it's a new feature)?
  • It doesn't change or implement public features.
• Did you write a regression test to reproduce the bug (if it's a bug fix)?
  • I'm pushed this commit to Fix vocab-tools generateCloner bug #541 too, so I think whether its CI pass or not can be treated as naive test.
• Did you write some tests for this change (if it's a new feature)?
  • It doesn't change or implement public features.
• Did you run mise test on your machine?

Additional notes

In the long term, it would be good to introduce a monorepo tool, but for now, it will be useful for CI stability.

[moreal]

47cd266 commit might be a better solution.

[moreal]

I believe this pull request can be replaced by #543