fix(spec): Conventions §Artefact-tiers — add re-frame2-resources (016) row

[mike-thompson-day8]

The drift

spec/Conventions.md §"Artefact tiers" — the per-feature table — is the canonical "what are the per-feature optional-capability artefacts" enumeration (machines / flows / routing / http / ssr / schemas / epoch). It omitted re-frame2-resources (Spec 016), even though resources is treated as a named per-feature artefact in every other location:

• spec/README.md Capability layer lists 016-Resources as an optional capability.
• spec/Ownership.md names day8/re-frame2-resources (post-v1 optional artefact) on the 016 row.
• spec/Conventions.md reserved-keys tables already mark :rf.runtime/resources + :rf.runtime/work-ledger + :rf.runtime/mutations as the post-v1 Resources artefact.
• spec/001-Registration.md marks the :resource registry kind late-bound by day8/re-frame2-resources.

So the table — the single complete per-feature enumeration — was the one place out of step.

Resolution — bead's preferred option (a)

The table caption says the CLJS reference's published artefact set; resources is post-v1 (not yet published). Rather than leave it out (which keeps the table incomplete) I took the bead's preferred option (a): include post-v1 artefacts with an explicit "(post-v1)" annotation so the table stays the single complete enumeration.

Changes (5 insertions / 2 deletions, table-only):

• New row re-frame2-resources (post-v1) → 016, annotated as the richest per-feature artefact: it alone contributes three runtime subsystems (:rf.runtime/resources / :rf.runtime/work-ledger / :rf.runtime/mutations) and three registrar kinds (:resource / :mutation / :resource-scope). Facts cross-checked against Ownership's 016 row and 001-Registration's :resource / :mutation / :resource-scope rows.
• Caption tweak — dropped "published" (no longer strictly accurate with a post-v1 row) and added a short clause noting the per-feature tier additionally enumerates post-v1 artefacts, tagged inline.
• Per-feature note — one line explaining the (post-v1) convention and that other post-v1 capabilities follow it as their coordinates settle.

Stories (007) stays a note

007-Stories is also post-v1, but unlike resources it has no fixed artefact coordinate in Ownership (just "post-v1"). Under option (a) Stories would need a coordinate decision first — out of scope here. So resources gets the annotated row; Stories is covered by the one-line note ("e.g. Stories (007) is a post-v1 capability without a fixed artefact coordinate yet, so it earns a row here once one is assigned") — no Stories coordinate invented.

Secondary (glossary) — skipped

I skipped the optional "optional capability = per-feature artefact" glossary line. The equivalence is not a triangulation the reader is forced to make at this point in the doc, and adding it would read as gold-plating on a hot-zone surgical edit. The new note already ties "post-v1 capability" → "earns a row," which is the only place the equivalence was load-bearing for this change.

Out of scope (untouched)

The downstream count inconsistencies (setup-skill "ten artefacts", "6 per-feature splits" phrasing in late-bind / bundle-isolation) are separate beads — not chased here.

Gate results

Ran the doc/spec gates the repo wires for spec/ changes (per .github/workflows/docs.yml), all PASS:

• check_doc_slugs.py (link target + anchor validator) — self-test + live: PASS
• check_runtime_subsystem_grading.py — self-test + live: PASS
• check_inject_cofx_residue.py — self-test + live: PASS
• check_retired_composition_vocab.py — self-test + live: PASS
• check_retired_image_keys.py — self-test + live: PASS
• check_ep_status_sync.py — PASS
• Staged spec/ → docs/spec then mkdocs build --strict — PASS (exit 0; only pre-existing INFO-level relative-link notes about examples/ and the-mayor-method/, none from this edit; no WARNING fired).

Table-only edit; no CLJS / spec-conformance gate keys off Conventions table content.

Refs: rf2-9ck5il