Ratchet coverage thresholds, refresh handoff, drop finished todo files

Coverage has grown well past the configured gates, so they no longer catch
regressions. Raise them to just under the measured values, per the standing
"regression ratchet, not a target" rule:
- frontend lines/statements 80 -> 95, functions 85 -> 89 (measured 96.52/91.08)
- frontend branches held at 85 (measured 86.78, the weakest metric)
- backend CI --cov-fail-under 50 -> 75 (measured 81)

Add NextSteps260914.md as the handoff for the next session. Delete todo.txt,
todo_260912.txt and todo-backend.txt, all fully done; NextSteps260913.md is
kept as history.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RXMhc9nT6sNz1SL4j8dznL

.github/workflows/ci.yml

@@ -33,8 +33,8 @@ jobs:
         run: uv run ruff format --check .
 
       - name: Test with coverage
-        # Target is 90; raise toward it as tests land (tips rules: 40→50).
-        run: uv run pytest --cov=app --cov-report=term-missing --cov-fail-under=50
+        # Measured ~81%; set threshold 1-2pp below to catch regressions.
+        run: uv run pytest --cov=app --cov-report=term-missing --cov-fail-under=75
 
   # ---------------------------------------------------------------------------
   # Frontend: ESLint + Vitest with 90% coverage gate