Run tests with file watcher

[stefnotch]

I wanted a Deno testing experience that's closer to Vitest's UI or Wallaby.js. As in

• See the test results inline in the IDE
• Re-run the tests and update the IDE's UI when saving

This now works 🎉
https://github.com/user-attachments/assets/b283b1e1-528f-47d5-b436-86c546d0d3c7

(The execution speed is very slow, because I have not compiled Deno in release mode)

To do so, I had to

• Bump the required VSCode version. This lets us take advantage of VSCode's "continuous run".
• Update the LSP to support these features. WIP - feat(lsp): Run tests with file watcher deno#27762
• Recreate the TestRun object when the tests are being restarted. This is because VSCode does not support changing the status of a test from failure back to success.
• Learn about a VSCode peculiarity Continuous test run: Two buttons to cancel, only one works microsoft/vscode#172205

Fix #1132

[CLAassistant]

All committers have signed the CLA.

[stefnotch]

Related PR with questions has been opened denoland/deno#27762

[nayeemrmn]

Does this make a difference to the currently resolved version and is it essential? I'd like to avoid it because the TS plugin API communication has been brittle in the past

[stefnotch]

Yes, this bit is essential. It's what is required to get request.continuous and the final argument in

new vscode.TestRunRequest(
              runData.request.include,
              runData.request.exclude,
              runData.request.profile,
              true,
            ),

Without that, we don't have the typings for the continuous run feature in vscode.

I believe that 1.77.0 is also the minimal version bump to get access to said feature.

[nayeemrmn]

LGTM

[stefnotch]

Thank you for merging this!

Would you have any suggestions regarding the Deno LSP PR that implements the internal functionality for this?
denoland/deno#27762

One of my biggest questions was: There's fair bit of code duplication im Deno. The same sort of "run normal" vs "run in watch mode" logic is duplicated here:
run_tests
https://github.com/denoland/deno/blob/57dd66ec3dae2a0fe376f6a43c476dfade421c84/cli/tools/test/mod.rs#L1555
run_tests_with_watch
https://github.com/denoland/deno/blob/57dd66ec3dae2a0fe376f6a43c476dfade421c84/cli/tools/test/mod.rs#L1647

exec
https://github.com/denoland/deno/blob/57dd66ec3dae2a0fe376f6a43c476dfade421c84/cli/lsp/testing/execution.rs#L216

run_benchmarks
https://github.com/denoland/deno/blob/57dd66ec3dae2a0fe376f6a43c476dfade421c84/cli/tools/bench/mod.rs#L426
run_benchmarks_with_watch
https://github.com/denoland/deno/blob/57dd66ec3dae2a0fe376f6a43c476dfade421c84/cli/tools/bench/mod.rs#L492

Now I did just "exec_with_watch" to the LSP and call it a day, but I feel like there might be a better approach. Any recommendations?