Raise e2e maxConcurrency back up (currently pinned to 2)

apps/os/e2e/vitest.config.ts runs the e2e suite with sequence.concurrent

• maxWorkers: 4 + maxConcurrency: 2 (peak ~8 concurrent tests). 2 is a deliberate stopgap: higher values overloaded the deployed preview slot — maxConcurrency: 6 (peak ~24) hit "Durable Object storage operation exceeded timeout", and 3 (peak ~12) still failed a timing-sensitive teardown test. The slot, not the runner, is the bottleneck: every e2e test creates a project (a whole DO chain), so concurrency multiplies cold/near-cold creates against one slot.

Goal: get back to ~6+ concurrent (the itx suite is 38 tests in one file and dominates the vitest lane) without slot overload. Approaches, likely combined:

• Split the itx monolith (e2e/vitest/itx.e2e.test.ts, 38 tests, ~287s sequential) into several files. File-level parallelism (maxWorkers) then speeds it up at a safe per-file concurrency — the clean way to get speed without cranking maxConcurrency.
• Fix cold-slot create latency ([[project_preview_e2e_speedup]] and tasks/os-cold-create-latency.md — someone is already on the 30-90s cold OAuth-callback / create-saga issue). Faster, more robust creates raise the concurrency ceiling directly.
• Consider a bigger/reserved preview slot, or pre-warming more of the create path at deploy.

Validate each bump with a preview dispatch and watch for "DO storage exceeded timeout" / timing-teardown flakes.