S2 — Rewrite the physical-architecture standard's directory tree (closes C-84) (Epic #240)

[Polichinel]

Part of epic #240. Blocked by S1 (#241) — both stories edit
docs/standards/physical_architecture_standard.md. S1 corrects the layering paragraph at :48-49 and
the Circular Dependency Guard at :67; this story rewrites the tree those lines sit under. Running
them in parallel conflicts, and running this one first ships the file half-corrected.

The problem

docs/standards/physical_architecture_standard.md is 78 lines, undated and unversioned, and its
authoritative directory tree (:30-44) is a pre-implementation design sketch that was never revised
after the package was built.

Measured against find src -name '*.py' (13 modules across 3 packages):

Claim in the tree	Reality
weight_frame.py "(anticipated)"	never shipped
mask_frame.py "(anticipated)"	never shipped
target_frame.py "(anticipated)"	shipped in v1.0.0, has its own CIC
metadata.py	missing from the tree
_typing.py	missing from the tree — and it has the highest fan-in in the leaf (8)
conformance/	missing from the tree — the one surface whose primary caller is another repository (ADR-016)
views_frames_summarize/	absent entirely
views_frames_reconcile/	absent entirely

Eight of thirteen real modules, one of three packages, two phantom files.

Why it matters

This is the document a contributor is pointed at to answer "where does this file go?", and ADR-007
treats agents as untrusted contributors who read exactly these documents. It is also the operational
companion to ADR-002 — the one that states the one-concept-per-file rule that
tests/test_import_enforcement.py::test_one_concept_per_file enforces.

Being undated is the compounding problem: every CIC carries a **Last reviewed:** line and all
eight are current (2026-06-24 to 2026-07-31). This standard has no such marker, so its staleness is
invisible to inspection and to validate_docs.sh.

Work

• Rewrite the tree at :30-44 to the three shipped packages and thirteen real modules. Derive it
  from find src -name '*.py' | sort, not from memory.
• Drop weight_frame.py and mask_frame.py. If the intent to add frame types should survive,
  state it as prose ("new frame types are additive MINOR, ADR-018") rather than as phantom files
  in an authoritative tree.
• Un-mark target_frame.py as "(anticipated)".
• Add metadata.py, _typing.py and the conformance/ subpackage, each with the same
  one-line-purpose + Category annotation the existing entries carry.
• Add views_frames_summarize/ and views_frames_reconcile/ — at minimum as package-level
  entries with their charters (ADR-017, ADR-023) and the note that siblings never import each
  other.
• Add a **Last reviewed:** YYYY-MM-DD line at the top, matching the CIC convention, so the next
  drift is detectable.
• Update C-84 in the register to RESOLVED. State what was verified — e.g. the module list was
  diffed against find src -name '*.py' — not what was intended (register C-77).

How we know it is done

• Every path in the tree exists, and every src/**/*.py module appears in the tree. Check
  mechanically:

  find src -name '*.py' | sed 's|src/||' | sort > /tmp/actual.txt
  # then confirm each appears in the standard
  

• The document carries a **Last reviewed:** date.
• No "(anticipated)" marker refers to something that shipped.
• bash docs/validate_docs.sh passes.
• The layering paragraph and Circular Dependency Guard from S1 are still correct — this story must not
  regress them.

Notes on scope

Do not expand this into rewriting the standard's rules (the one-concept-per-file rule, the naming
convention, the Category taxonomy). Those are accurate and enforced by tests. This story fixes the
tree and the date, nothing else.

Relevant files

docs/standards/physical_architecture_standard.md (the only file changed, plus the register) ·
reports/technical_risk_register.md (C-84) · tests/test_import_enforcement.py::test_one_concept_per_file
(the test that enforces the rule this document states)