Establish the deterministic headless desk test harness before feature screens

[clintecker]

Problem

A TUI can appear correct while key bindings, focus, async messages, resizing, and subprocess transitions are wrong. Waiting until screens exist to add a harness would accumulate untestable UI behavior.

Direction

Add a reusable headless Textual harness using App.run_test() and Pilot, fixed viewport fixtures, injected read models/processes, semantic widget queries, and active-signal assertions. Build it alongside the first app shell, before DESK and RUN feature work.

Dependencies

• Optional package/entry boundary: Package an optional Textual desk with one canonical press desk entry #104 — Package an optional Textual desk with one canonical press desk entry #104
• General pytest foundation: Establish the pytest harness and migrate press selftest into reusable tests #78 — Establish the pytest harness and migrate press selftest into reusable tests #78
• Composable factories: Build a composable book and artifact factory for isolated test scenarios #83 — Build a composable book and artifact factory for isolated test scenarios #83

Invariants

• Tests contain no sleeps, wall-clock deadlines, real tool probes, or terminal-dependent timing.
• Pilot interactions assert state, return values, refusals, and terminal outcomes—not only screenshots.
• Every async transition has an observable completion signal.
• Fixed 80×24 and expanded viewports are first-class fixtures.
• Snapshot tests supplement semantic assertions and never replace them.

Acceptance

• Harness helpers construct the app against a composable book fixture and smart fake adapters.
• Tests demonstrate key presses, clicks, focus, resize, worker completion, refusal, and failure paths.
• Failures include the active screen, focused widget, model state, and captured child output.
• A bounded snapshot policy defines review/update rules and stable SVG normalization.
• The harness is documented for contributors and linked from the invariant ledger.

Links

• Official Textual testing guide: https://textual.textualize.io/guide/testing/
• Pilot API: https://textual.textualize.io/api/pilot/
• Milestone: https://github.com/clintecker/press/milestone/11