Splitter: structural rules + tests; regenerate as 14-part hierarchy

[postpersonality]

Summary

Three deviations were producing spurious or mis-nested folders under the
naive H1/H2 split:

• False H1s from table rows (# | Block | FPF U.Type | …) — two
  markdown table rows in Part F were parsed as section headings, creating
  empty block-fpf-u-type-… and block-base-concept-scale-map folders.
• Empty H1 wrappers — # Part E and # Part F are top-level
  wrappers whose only child is another # H1 (Section E-I,
  Cluster F.I), producing two empty folders next to two real ones.
• Misplaced clusters — # Cluster A.IV.A and # Cluster A.V are
  semantic children of Part A but authored as # H1, so they appeared
  as siblings of Part A instead of nested under it.

This PR adds three rules to scripts/split_spec.py, covers them with
unit tests, regenerates sections/, and updates SKILL.md and the READMEs
to match.

The splitter remains a representation morphism: pattern text in leaf
files is preserved verbatim. Only the directory structure and the
auto-generated _index.md TOCs change.

Result

	Before	After
Top-level Parts	20	14
Total directories	20	16
Total .md files	245	241
Empty wrapper folders	2	0
Spurious folders (table)	2	0
Nested clusters	0	2

The two new nested folders sit under Part A: cluster-a-iv-a-… (22 leaf
files for the Signature Stack) and cluster-a-v-… (31 leaf files for
the Constitutional Principles).

Commits

Each commit is atomic and reviewable independently.

1. chore: bump FPF submodule and regenerate sections (semio R1+ upstream)

Pure upstream sync — no script changes. FPF submodule bumped from
a25a0a7 to b18acde (terminology cleanup in E.8/E.9/E.19 plus
accumulated semio R1+ changes: SD→AUD rename, new patterns under A.6,
A.19, C.11, C.22.1, C.26 family, E.11, I.2, J.4). Sections regenerated
with the existing splitter so the structural diff is purely upstream
content evolution.

2. feat(splitter): structural rules with tests; regenerate as 14-part hierarchy

Splitter additions:

• Rule A — heading regex now ignores ^# \| lines (table rows
  pretending to be H1).
• Rule B — empty H1 wrappers fold into the next H1's preamble; the
  merged folder takes the wrapper's title (canonical Part name).
• Rule C — H1s whose pattern ID (A.IV.A, A.V, …) matches the
  previous Part's letter become subfolders of that Part instead of
  top-level peers.

Plus scripts/test_split_spec.py with six unittest cases — three for
the new rules, three as regression cases ensuring current correct
behaviour is preserved (title-page-with-content does not merge with the
next H1; normal H1 with H2 children is unchanged; H1 with preamble but
no H2 children does not merge with an unrelated peer).

.gitignore added for __pycache__/ and *.pyc.

sections/ regenerated under the new rules — same content, fewer
folders, 2 nested clusters under Part A.

3. docs: update SKILL.md and READMEs for 14-part nested structure

• SKILL.md thinking-verb router: renumber section references
  (Part B 07→05, Part C 08→06, Part D 09→07, Part E 11→08, Part F 13→09,
  Part G 16→10, Part K 20→14); route A.IV.A and A.V via Part A.
• SKILL.md Step 2: note that nested clusters require descending
  into the cluster's own _index.md after the parent Part's.
• SKILL.md Section INDEX: collapse from 20 rows to 14, fold A.IV.A
  and A.V into Part A's row with linked sub-cluster descriptions.
• README.md / README-RU.md: rewrite the abstract (16 dirs, 241
  files), the structure block (showing nested clusters), and the
  Sections table.
• FPF-SKILL-UPDATE-GUIDE.md: refresh router-row example and add a
  note about nested-cluster path syntax.

Test plan

• python3 scripts/test_split_spec.py — 6/6 PASS on the new rules
  and regression cases.
• python3 scripts/split_spec.py on the bumped submodule produces
  14 top-level Parts + 2 nested clusters; output reports
  "Wrote 16 directories, 241 files".
• git status after regeneration is clean — splitter is
  idempotent.
• All inbound _index.md references in SKILL.md resolve to
  existing folders.

Reviewing tips

• Commit 1 is a large mechanical diff (bumped submodule + regenerated
  files). Skim the FPF submodule pointer change and read the spec
  diff at the upstream repo for content review.
• Commit 2 is the substance — read scripts/split_spec.py and
  scripts/test_split_spec.py. The sections/ diff is consequential
  and reproducible by re-running the splitter.
• Commit 3 is small and focused on navigation.

[gemini-code-assist]

Code Review

This pull request updates the FPF specification structure, introducing a hierarchical organization with sub-clusters and improving the split_spec.py script to handle these changes. Several documentation files were updated to reflect these structural improvements. I have identified a systematic search-and-replace error where U. was incorrectly replaced by s. in sections/04-part-a-kernel-architecture-cluster/cluster-a-v---constitutional-principles-of-the-kernel/04-a-10---evidence-graph-referring.md, and a logic issue in scripts/split_spec.py where an empty 'Contents' header is generated for sections containing only sub-clusters.