address architecture review findings and add review document

Fixes from architecture review (4 critical, 10 warnings):

Critical:
- Fix selectNode/setNode docs to accurately describe prop-key
  navigation behavior including array support and prop-key writes
- Document RenderContext/Density exported types in host-config
- Resolve ADR dual status ambiguity with clarifying note in README
  (frontmatter status = editorial, body Status = decision)
- Effect types already addressed in prior commit

Warnings addressed:
- Add Fragment re-export note to jsx-runtime section in
  build-distribution
- Document childCtx/transformCtx helper functions in transforms.md
- Document render() accepting non-root UNode in host-config
- Add Value.Hash re-entrancy constraint to reconciler.md
- Add true-passthrough constraint and h('root') special case
  to element-factory constraints
- Add _idCounter bundling caveat note

Review document added at docs/reviews/architecture-review-2026-05-18.md
with full findings, source verification table, and recommendations.

docs/architecture/reconciler.md

@@ -244,6 +244,7 @@ Host implementations that currently only support mount-only rendering will start
 - **`key` is a reconciler concern, not a prop** — `key` is extracted by `h()` and never passed to components or hosts. See [ADR-004](decisions/004-key-as-first-class-field.md).
 - **Commit order is parent → child, top-down** — this ensures parent state is consistent when child updates fire.
 - **TypeBox optimizations are incremental** — the reconciler is correct without them. Each optimization layers on top.
+- **`Value.Hash` is not re-entrant** — it uses a global mutable accumulator. It must not be called from within a `computed` or `effect` that is itself triggered by a hash comparison. Hashes must be computed outside reactive computations, during the commit phase.
 - **Phase 5 (flowgraph host configs) belongs in `@alkdev/flowgraph`** — ujsx stays generic. Flowgraph is a consumer, not a modification.
 
 ## Open Questions