Redesign TODO test output: separate axis with pending/resolved

[gfx]

Summary

• TODO tests (#[TODO], #![TODO]) are now reported on a separate axis from pass/fail, inspired by TAP's TODO directive
• Pending (body traps): expected — the feature is still unimplemented. Does not affect exit code.
• Resolved (body passes): hard failure (exit code 1) — enforces removing #[TODO] when the feature works.
• New summary format: N passed, N failed; N todo (M resolved)
• TODO summary section at the end lists all TODO tests with file path and status
• Formal specification added to docs/spec.md (Test Outcome Model section)

Motivation

Previously, TODO tests were "expected to fail" — if a compiler fix caused a TODO test to start passing, it showed up as a test failure, which was confusing and made compiler development harder. Now, newly-passing TODO tests appear as "resolved" on the TODO axis instead.

Output example

Running tests in math_test.wado... (compiled in 50ms)
  ✓ addition (2ms)
  · future_feature # TODO (1ms)
  ✓ now_works # TODO resolved (2ms)
    remove the #[TODO] attribute

TODO tests (2):
  · pending   math_test.wado — future_feature
  ✓ resolved  math_test.wado — now_works

1 TODO test(s) resolved — remove the #[TODO] attribute

1 passed, 0 failed; 2 todo (1 resolved) (0.12s)

Test plan

• cargo check --bin wado passes
• cargo check --test e2e -p wado-compiler passes
• Unit tests for test name parsing pass
• mise run on-task-done (running)

https://claude.ai/code/session_01Dt11N9EJkrwhjhC7PhMLQG