iter effect-doc-honesty: make the effect-system documentation true

Standalone documentation-honesty tidy (no language/checker/codegen
change; `ail check`/`run` byte-unchanged by construction). Corrects
three false effect-system claims the effect-subsystem recon surfaced,
plus two satellite mentions, guarded by a new doc-presence pin:

- DESIGN.md Decision 3: removed the "row-polymorphic (`![IO | r]`)"
  claim (no EffectRow / row variable exists in any crate — effect
  sets are a flat, unordered, closed set unified by set-equality);
  reconciled "`IO` and `Diverge` are wired up" to IO-only +
  `Diverge` reserved/unimplemented (zero code in any crate),
  modelled on Decision 4's reserved-refinements precedent.
- DESIGN.md "What is not (yet) supported": "IO and Diverge ops"
  bullet -> IO-only + Diverge-reserved.
- ast.rs Term::Do doc-comment: "resolved against the effect-handler
  table at link time" -> the real mechanism (typecheck lookup in
  Env::effect_ops + literal lower_effect_op codegen match; no
  handler table, no link-time resolution).
- form_a.md:226 + rule 3, and main.rs merge-prose CONTRACT example:
  Diverge mentions reconciled in lockstep.
- new crates/ailang-core/tests/effect_doc_honesty_pin.rs: 4 tests,
  fiction-absent + corrected-anchor-present, single-line wrap-robust
  substrings.

cargo test --workspace 600 -> 604 (4 new pin tests, 0 regressions);
design_schema_drift / spec_drift / schema_coverage stay green,
empirically confirming none scans the effect-prose region.

bench/orchestrator-stats/2026-05-16-iter-effect-doc-honesty.json

@@ -0,0 +1,13 @@
+{
+  "iter_id": "effect-doc-honesty",
+  "date": "2026-05-16",
+  "mode": "standard",
+  "outcome": "DONE",
+  "tasks_total": 6,
+  "tasks_completed": 6,
+  "reloops_per_task": { "1": 0, "2": 1, "3": 0, "4": 0, "5": 0, "6": 0 },
+  "review_loops_spec": 0,
+  "review_loops_quality": 0,
+  "blocked_reason": null,
+  "notes": "Task 2 took 1 implementer-internal repair re-loop (not a spec/quality review loop): the plan's verbatim Task-2 replacement body soft-wrapped 'flat, unordered,\\nclosed set of effect names' mid-phrase, contradicting Task-1's single-line pin substring and the plan's own self-review item 6. Resolved by re-wrapping the soft-wrap column only (wording byte-identical to plan intent; pin unchanged). Recorded as a planner self-review item-3 cross-check gap."
+}