fix(e2e): run server tests sequentially

[SulimanAbdulrazzaq]

Summary

Closes #225.

Each E2E test file starts a full OpenFox server. Starting up to 12 of these servers concurrently can exhaust local CPU and memory long enough for createTestServer() to exceed the existing 10-second hook timeout.

Run server-backed E2E files sequentially, matching the already reliable CI worker policy. This addresses the resource contention directly without increasing test or hook timeouts.

A regression assertion now protects the worker policy.

Validation:

• Reproduced the default-worker startup timeout cascade
• Confirmed four workers still fail with the same hook timeouts
• e2e/timeouts.test.ts: 4 tests passed
• E2E TypeScript check passed
• Prettier passed for both changed files
• Full sequential E2E run was capped after approximately eight minutes under concurrent batch load and did not produce a final result

AI-Enhanced Development

• AI Models: GPT-5.6 (Codex)

Cache Impact

• No

[co-l]

Thanks for digging into this — the diagnosis (parallel full-server startups blowing the hook budget) is spot on.

We went with a gentler fix on develop instead of forcing sequential for everyone: the worker count is now configurable per machine via an env var, with the old defaults preserved.

• Default: 12 workers locally, 1 on CI (unchanged).
• Override per machine: OPENFOX_E2E_MAX_WORKERS=<n>
  • e.g. OPENFOX_E2E_MAX_WORKERS=4 npm run test:e2e
• Invalid values (non-integer, <1, garbage) fall back to the default.

The concern with merging this PR as-is was that hardcoding maxWorkers: 1 — plus the new regression test asserting it — would permanently remove the ability to parallelize for everyone, including machines that handle 12 fine. This way the knob stays.

Closing this one; if your machine still can't cope with the env var set low, reopen and we can look at a smarter default (e.g. derived from CPU count).