feat(610): split the decision corpus into one YAML-frontmatter file per record

168 records -> docs/decisions/records/<area>/<topic>.md (163 active, 23 dirs) and
docs/decisions/archive/<area>/<topic>.md (5 archived). The filename IS the key,
so one-active-record-per-key becomes a filesystem property rather than a
validator check, and supersession becomes a `git mv`.

WHY: the monolith was a concurrency problem before an aesthetic one. A
3,900-line append target made parallel sessions collide -- PR #605 and PR #614
both hit append-vs-append conflicts during routine rebases, and hand-resolving
those inside the corpus is exactly the operation the rationale-rewrite guard
exists to police.

HOW IT IS VERIFIED: a ~170-file diff cannot be meaningfully read, so correctness
does not rest on reading it. The parser was taught BOTH formats first, so the
body-diff guard parses the old form at the merge-base and the new form at head --
the migration validates itself, no bypass. The proof is a field-level equivalence
harness: 168 records before and after, zero lost, zero gained, zero field
mismatches, zero rationale bodies differing. Reviewers should scrutinise the
harness; it is the actual evidence.

What measuring caught that reading would not have:

- ~500 lines sit OUTSIDE any record -- decisions.md's lifecycle schema and each
  topic file's preamble, mostly the only copy. Source files are kept and
  stripped, never deleted. They also cannot be filed per-area: topic files hold
  several areas and 4 of 23 areas span several files.
- Archive discovery was a non-recursive glob; after the split it found ZERO
  archived records, surfacing as four bogus "supersedes points to unknown key"
  errors rather than an obvious failure.
- ~32 live docs point into the corpus BY DATE, which the split dangles. Each
  stripped file now ends with a generated "Records formerly in this file" index,
  which also rescues the identical breadcrumbs in old issue comments.
- decisions.md's "In this file:" list was 97 same-file anchor bullets that the
  split makes WRONG, not merely stale. Dropped; the generated index replaces
  them with links that resolve.

The equivalence harness now runs against a checked-in FIXTURE, not the live
corpus. The earlier version migrated the real tree, which made it a one-shot:
the moment the migration landed there was nothing left to move and the tests
failed for reasons unrelated to the code. A fixture keeps them testing the
SCRIPT rather than the repo's current state.

Keys preserved verbatim, warts included: `sched` (12) and `scheduling` (1) remain
two directories for one concept. Renaming a key is not a move -- it changes
identity, breaks the equivalence proof, and invalidates MemPalace's per-key
drawers. Taxonomy normalisation is separate work.

refs #610

scripts/migrate_decisions_split.py

@@ -13,15 +13,19 @@ Two properties this script exists to guarantee, both checked by
 1. **Nothing is rewritten.** Every record's rationale body is copied byte-for-byte, and every
    metadata field round-trips unchanged. The migration is a MOVE, so a field-level before/after
    comparison is a complete correctness proof.
-2. **No prose is lost.** 506 lines of the corpus are NOT inside any record — file preambles that
-   explain why a topic file exists, cross-reference notes, and decisions.md's lifecycle schema
-   header. Those files are kept and stripped to exactly that prose, never deleted. (They cannot be
-   filed per-area: topic files hold several areas, and 4 of 23 areas span several files.)
+2. **No prose is lost.** A few hundred lines of the corpus are NOT inside any record — file
+   preambles that explain why a topic file exists, cross-reference notes, and decisions.md's
+   lifecycle schema header. Those files are kept and stripped to exactly that prose, never deleted.
+   (They cannot be filed per-area: topic files hold several areas, and 4 of 23 areas span several
+   files.) The one thing deliberately dropped is same-file anchor bullets, which the split makes
+   *wrong* rather than merely stale — the generated index replaces them with links that resolve.
+   Run with `--dry-run` for the current counts rather than trusting a number written here.
 """
 
 from __future__ import annotations
 
 import argparse
+import re
 from pathlib import Path
 
 import yaml
@@ -121,8 +125,19 @@ def _moved_index(records: list[dl.Record]) -> str:
     return "\n".join(lines)
 
 
+# A bullet whose link is a bare same-file anchor, e.g. `- [2026-07-10 — Title](#2026-07-10--title)`.
+# decisions.md's "In this file:" index is ~97 of these. Once the records move out, every one of them
+# points at a heading that no longer exists in that file — and the generated "Records formerly in
+# this file" section replaces them with links that actually resolve. Anything else keeps its bullet.
+_SAME_FILE_ANCHOR_BULLET = re.compile(r"^\s*[-*]\s*\[[^\]]*\]\(#[^)]*\)\s*$")
+
+
 def _preamble(path: Path, records: list[dl.Record]) -> str:
-    """Everything in `path` that is not inside a record — kept, never deleted."""
+    """Everything in `path` that is not inside a record — kept, never deleted.
+
+    Two exceptions are dropped rather than kept, because the split makes them *wrong* rather than
+    merely stale: same-file anchor bullets (see above).
+    """
     text = path.read_text(encoding="utf-8")
     lines = text.splitlines()
     drop: set[int] = set()
@@ -130,10 +145,16 @@ def _preamble(path: Path, records: list[dl.Record]) -> str:
         start = r.lineno - 1  # heading line, 1-indexed in Record
         for n in range(start, start + 1 + len(r.body.splitlines())):
             drop.add(n)
-    kept = [ln for n, ln in enumerate(lines) if n not in drop]
-    while kept and not kept[-1].strip():
-        kept.pop()
-    return "\n".join(kept)
+    kept = [ln for n, ln in enumerate(lines) if n not in drop and not _SAME_FILE_ANCHOR_BULLET.match(ln)]
+    # collapse the blank runs left behind where a bullet list used to be
+    out: list[str] = []
+    for ln in kept:
+        if not ln.strip() and out and not out[-1].strip():
+            continue
+        out.append(ln)
+    while out and not out[-1].strip():
+        out.pop()
+    return "\n".join(out)
 
 
 def plan() -> tuple[list[tuple[dl.Record, Path]], dict[Path, str], dict[Path, int]]: