Replace module mocks with Effect dependency injection #172

problem

The anti-slop migration in #171 exposes 50 vi.mock calls across 23 test files. Whole-module mocks couple tests to import topology, can drift from production behavior, and make refactors noisy.

direction

Replace the internal workflow mocks with a few real dependency seams. Prefer Effect services and layers where the production code already uses Effect:

1. Import workflow: audio backend, provider lookup, downloads, metadata decoding, and analytics.
2. Share workflow: share client, publication capability storage, presence, cover processing, notifications, and analytics.
3. Export/workspace workflow: download execution, system-failure reporting, audio backend, notifications, and analytics.
4. UI-library mocks: render the real shadcn/DnD behavior where practical, or test the underlying controller through its interface. Do not invent production interfaces solely to replace a third-party UI mock.

Each seam should have both a live adapter and a test adapter. New tests should exercise behavior through the workflow interface; remove the superseded shallow tests instead of layering duplicate coverage.

acceptance criteria

• No vi.mock, vi.doMock, or equivalent whole-module mocks remain in tests/**.
• Import, share, export, and workspace behavior retain focused coverage through real interfaces.
• Test adapters are typed and faithful to the live interface.
• Obsolete implementation-coupled tests are removed.
• anti-slop/no-module-mocking can be promoted from warning to error.
• Lint, typecheck, unit/server tests, and the relevant browser flows pass.

non-goals

• Rewriting unrelated production modules.
• Adding pass-through interfaces with only one adapter.
• Preserving tests whose only value is asserting implementation details.