…ate timeout (#4087)

## Problem

`showcase/scripts` test runs intermittently fail with birpc
`onTaskUpdate` timeouts, caused by an upstream bug in vitest 3.x's
subprocess reporter transport
([vitest-dev/vitest#8164](vitest-dev/vitest#8164)).
Fix landed upstream in
[#8297](vitest-dev/vitest#8297), shipped in
v4.0.0-beta.4+.

## Why this is safe

- **Blast radius is this one package.** Root `package.json` already
declares `"vitest": "^4.1.3"`. The only holdout was
`showcase/scripts/package.json` at `"vitest": "^3.0.0"`. All 10
subprocess-spawning test files in the monorepo live in
`showcase/scripts/__tests__/` — zero elsewhere — so this is the only
package actually affected by the bug.
- **Zero vitest API breaks in the diff.** Spot-checked: no `vi.mock`, no
`expect.extend`, no `.concurrent`, no snapshots, no coverage config, no
custom reporters. Only `vi.spyOn` (unchanged in v4) and standard
`describe`/`it`/`expect`.
- **Pool config unchanged.** `pool: 'forks'` + `fileParallelism` behave
identically in v3 and v4.
- **Lockfile diff is minimal.** Only `showcase/scripts` import
resolution changes + a new `@vitest/[email protected]` snapshot. No other
package was edited; no other package's resolution changed. vitest 3.2.4
and 2.1.9 continue to resolve side-by-side for the packages that still
pin them.

## Verification

3 consecutive `pnpm nx run @copilotkit/showcase-scripts:test
--skip-nx-cache` runs on Node 20.20.2:

| Run | Result | Wall time | Tests |
|-----|--------|-----------|-------|
| 1 | green | 34.45s | 1061/1061 |
| 2 | green | 29.64s | 1061/1061 |
| 3 | green | 29.46s | 1061/1061 |

Zero `onTaskUpdate` errors, zero unhandled exceptions across all three
runs.

## Test plan

- [ ] CI unit(20.x) green — this has been the red one on recent PRs due
to this exact bug
- [ ] CI unit(22.x) green
- [ ] CI unit(24.x) green

## Relationship to #4081

Complementary, not overlapping. #4081 splits / isolates the
showcase-scripts test suite as a workaround at the Nx/pool-config layer.
This PR fixes the root cause in the test runner itself. Once this
merges, the `onTaskUpdate` failures #4081 has been chasing should clear;
#4081 can then rebase on a clean main and its remaining work (the
split/isolation improvements) can be evaluated on its own merits rather
than as a bug workaround.

Vitest 4 lets us drop the workarounds that gated #5 (Windows continue-
on-error) and #4 (poolMatchGlobs federation override): birpc now
respects user-configured timeouts (vitest-dev/vitest#8297), the legacy
basic reporter is replaced by default+silent=passed-only.

Also bump @vitejs/plugin-vue 5.x → 6.0.6 in web-panel and desktop-app-
vue: vitest 4's vite peer requires vite ^6 || ^7 || ^8, plugin-vue 5
caps at vite ^6, conflict surfaces as ERESOLVE on web-panel install.

Affected: cli, web-panel, core-mtc, session-core, shared-logger,
core-{db,env,infra,config}, desktop-app-vue (incl. @vitest/coverage-v8
and @vitest/ui).

Vitest 4 removed test.poolOptions; forks/threads/etc are now top-level.
Also drop poolMatchGlobs from packages/cli/vitest.config.js — it routed
mtc-federation-governance-*-cli files to the threads pool to dodge
v3's hardcoded 60s birpc onTaskUpdate heartbeat (issue #4). Vitest 4
fixes the underlying RPC bug (vitest-dev/vitest#8297, shipped v4.0.0),
so the workaround is no longer needed; the file split done in
0114aec stays.

The "Unhandled Error: [vitest-worker]: Timeout calling onTaskUpdate" we
were seeing in CodSpeed CI is a known vitest 3.x bug
(vitest-dev/vitest#8164) where birpc's hardcoded 60s RPC timeout fires
when the worker thread is blocked by slow synchronous code — exactly
what happens under CodSpeed's Valgrind instrumentation.

The fix shipped in vitest 4.0.0-beta.4 (vitest-dev/vitest#8297) and is
now in stable 4.1.6. CodSpeed's vitest plugin already supports vitest 4
(peer: ^3.2 || ^4), and our existing Vite 7 + Node 20+ stack meets
v4's prerequisites.

Now that the RPC timeout is solved at the source, the bench-only
forks/singleFork/inflated-timeout workaround we added in eb902be is no
longer needed; reverted those vite.config tweaks to keep the configs
small and uniform across test and bench runs.

Also widens tooling tsconfig types so bench files can read process.env
(needed by the CodSpeed env guard in update-partial.bench.tsx, which
was already there but was previously masked).

Co-authored-by: Cursor <[email protected]>

CI flaked sporadically on test-engine with the unhandled error
`[vitest-worker]: Timeout calling "onTaskUpdate"`, even though every
test passed. Three prior deflake attempts (commits 884df74, 3e1a34a,
8dcd25f) addressed surrounding pressure but couldn't fix the root cause:
vitest 3.x's birpc dependency had a hardcoded 60-second DEFAULT_TIMEOUT
on inter-process RPC calls. Under CI load, an `onTaskUpdate` heartbeat
from a worker fork could land past 60s while the main reporter was
busy, and the stale promise would then reject right at teardown.

vitest 4 (PR vitest-dev/vitest#8297, shipped in 4.1.6) sets
`timeout: -1` on createBirpc — no heartbeat timeout. The flake is
structurally impossible on 4.1.7. Verified locally: 171/171 test files,
2926/2926 tests, no unhandled error.

Migration:
 - vitest.config.ts: `poolOptions.forks.maxForks` → top-level
   `maxWorkers`, drop the no-longer-relevant `minForks`. Comment
   updated to reflect the new root-cause understanding.
 - tsconfig.app.json: add `"node"` to types. Vitest 4 stopped pulling
   `@types/node` in transitively, so a couple of src/harness/__tests__
   files using `__dirname` / bare `'fs'` imports started failing
   typecheck. Adding node explicitly is the right fix.
 - src/hooks/__tests__/useInventoryDrag.test.ts +
   src/hooks/__tests__/useKeyboard.test.ts: `ReturnType<typeof vi.fn>`
   typed as `Mock<Procedure | Constructable>` in vitest 4 (vs the more
   permissive vitest 3 type), so it no longer assigns to a `() => void`
   parameter. Switched to the more specific `MockedFunction<() => void>`.

Side benefit: a future *real* RPC hang now never surfaces via the 60s
timer — it surfaces only via teardownTimeout (60s) or testTimeout (30s),
which is fine. We were already past the 60s window in CI and the
hardcoded value was hiding genuinely-fast tests behind a flaky failure
mode.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
Skip-Harness: tooling change — vitest dependency upgrade only, no product code touched

Vitest 3.x birpc caps worker RPC at 60s while subprocess-heavy tests use
120–180s testTimeout; long spawnSync blocks trigger false onTaskUpdate
timeouts in CI even with per-file threads. Vitest 4 aligns RPC with test
timeouts (vitest-dev/vitest#8297).

Co-authored-by: Cursor <[email protected]>