Add end-to-end gameplay flow tests for core colony interactions

[joryirving]

Summary

Add end-to-end gameplay flow tests that exercise real core player interactions instead of relying only on unit-like script checks.

Why

Windowstead now has enough interaction complexity that startup tests are not enough. Core flows like opening the menu, changing dock anchor, placing builds, saving/loading, and letting workers resume should be validated together.

Scope

Build a lightweight E2E harness that can run headlessly in CI and cover a few high-value flows such as:

• boot the game and open/close the menu
• switch anchors and verify the app remains usable
• enter build placement, place a structure, and cancel placement
• save, reload, and verify the colony resumes in a sane state

Implementation Guidance

• Work under tests/ and .github/workflows/test.yml
• Keep the number of E2E flows small and high-value at first; this issue is about establishing the path, not building an exhaustive matrix
• Prefer deterministic scripted interactions over brittle timing-heavy hacks
• Use this to cover real user-facing flows that have already regressed in prerelease testing

Acceptance Criteria

• CI runs at least one real end-to-end gameplay interaction flow
• The harness can cover menu, anchor, build-placement, and save/load behavior incrementally over time
• The initial E2E tests focus on high-value user flows rather than exhaustive coverage
• The project has a clear place to add future end-to-end regression scenarios