Replace Jest with Vitest 2

[birkskyum]

This PR replaces the test runner (unit/integration/build) from Jest to Vitest 2, which has built-in support for TS/ESM and the same syntax as Jest. This PR serves to help us evaluate the option, and to move forward faster if there's a decision to proceed at some point.

A key motivation for examining this change is that I've found it quite challenging to move these tickets forward:

• ECMAScript Modules #1595
• Tree shaking #977

Part of the struggle is that Jest only has experimental ESM support, and I've so far not managed to see an effect of the useESM setting in the ts-jest plugin used in this repo.

It's a continuation of

• Use platform consistently instead of jest done #4684

Relates to Rolldown, because vitest is downstream of vite, and thus eventually will swap internals to use rolldown:

• Future: Move bundling from Rollup to Rolldown #4669

Relates a bit to browser runner, because vitest is working on a browser mode:

• Future: Replace Puppeteer with Playwright #4949

Launch Checklist

• Confirm your changes do not include backports from Mapbox projects (unless with compliant license) - if you are not sure about this, please ask!

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 0% with 2 lines in your changes missing coverage. Please review.

Project coverage is 71.14%. Comparing base (d3f2bca) to head (28ec0ae).
Report is 4 commits behind head on main.

Files with missing lines	Patch %	Lines
src/source/tile.ts	0.00%	0 Missing and 1 partial ⚠️
src/util/web_worker_transfer.ts	0.00%	0 Missing and 1 partial ⚠️

❗ There is a different number of reports uploaded between BASE (d3f2bca) and HEAD (28ec0ae). Click for more details.

HEAD has 3 uploads less than BASE

Flag	BASE (d3f2bca)	HEAD (28ec0ae)
	9	6

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #4728       +/-   ##
===========================================
- Coverage   90.48%   71.14%   -19.34%     
===========================================
  Files         265      262        -3     
  Lines       38113    37819      -294     
  Branches     3166     1955     -1211     
===========================================
- Hits        34488    26908     -7580     
- Misses       2650     9638     +6988     
- Partials      975     1273      +298     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jasongitmail]

@birkskyum I took a look at the unit test config as you requested.

Looks good afaict. Coverage is in a separate command, fileParallelism is left at the default (true), etc.

If there is a particular file that consistently runs more slowly, using describe.concurrent() within it could help. And I'd carefully look at how beforeEach() is used and consider if beforeAll() would be sufficient.

I don't claim to be a Vitest expert though. I switched from Jest -> Vitest -> Bun, and experienced speed wins each time. But you're using newer versions of Jest and Vitest, so things may have changed. And these tests use JSDom, which I've never needed to use, which is slower than running tests on node.

The benefit of this PR, is not only does Vitest provide ESM support, it brings some future proofing because Vitest should eventually be rewritten in native code, given VoidZero's vision for a rust based end-to-end toolchain.

[birkskyum]

@jasongitmail , thank you very much for the review. It is a good idea to try to narrow down single files that might be impacting the numbers, and if beforeEach/All .concurrent() can help. I haven't looked into either of those, I just ported the tests directly, so it's likely there's something to gain.

You can run Vitest with Bun, and see a speedup? I haven't had much success with that. Unless the --bun flag is set, I believe bun just fires up node to run vitest. If you can run with this flag, please let me know, because then I have to update some tickets. If you can see a speedup running without the --bun flags, that's quite surprising to me given it's likely the same node process as not using bun, but they might of course have a trick.