fix(validator): C2.6 — finiteness hardening, Infinity can no longer vacuously clear the gate (closes R-2)

IR schema now refuses non-finite numbers (allow_inf_nan=False on quantity/
unit_cost/claimed_saving_nok) and non-finite or negative assumption-band
endpoints; json.loads accepts the bare Infinity literal, so the bundle seam
is tested directly. ModelMapContract rejects empty-string model ids
(min_length=1). check_turn_safety_net documented as a deliberately
unreachable belt under the range-bound debate loop.

18 new tests; detach-proven (re-allow inf/nan -> 5 red, drop min_length ->
2 red). Full gate: 365 passed, ruff/format/mypy clean; golden untouched.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/portfolio_optimiser_claude/loop.py

@@ -109,7 +109,12 @@ class DebateResult:
 
 
 def check_turn_safety_net(turns: int, max_rounds: int) -> None:
-    """The turn-count termination safety net ABOVE the round cap (§3 Step 3, §8)."""
+    """The turn-count termination safety net ABOVE the round cap (§3 Step 3, §8).
+
+    Under the ``range(max_rounds)``-bounded debate loop, ``turns`` never exceeds
+    ``2 * max_rounds``, so this net is structurally unreachable — it is a
+    deliberate belt that fires only if a refactor breaks the loop's own bound.
+    """
     if turns > 2 * max_rounds + 2:
         raise RuntimeError(
             f"debate turn-count safety net tripped: {turns} turns with max_rounds={max_rounds}"