volt: bound the tool_map drift CHECK (42.75s, was 107.93s and killed mid-leg)

The CHECK shipped with 6b7d750 ran the un-flagged repo gate: 107.93s wall,
82s of it the selftest-sweep leg. Inside the caller's harness it was killed
just after printing the section-4 header, so a green tree read as a failure.

Rebuilt as wizard-work/volt/CHECK-toolsmd-drift: per-leg timeout, the sweep
dropped from the budget (it is a wrapper over the cells graded directly here),
the leg tally asserted exactly, elapsed total printed. Non-vacuity measured by
three mutations on the real tree -- document drift (leg 1), uncommitted tool
drift (leg 2), index drift (leg 4) -- each rc 1, each restored clean.

wizard-work/volt/CHECK-toolsmd-drift

@@ -0,0 +1,58 @@
+#!/bin/sh
+# 2026-10-30 tool_map drift repair -- CHECK (bounded rebuild).
+#
+# The first version of this script ran `tools/verify.py` un-flagged, i.e. with
+# the selftest-sweep leg, and measured 107.93s wall on a clean tree. Inside the
+# caller's harness it was killed just after printing this section's header, so
+# the script read as a failure while every graded property in it was green
+# (the un-flagged gate, run by hand on the same tree, reads 9/9 and rc 0).
+# A CHECK that cannot finish inside its own harness is not a check.
+#
+# So each leg now has an explicit `timeout`, the sweep is out of the budget
+# (it is a wrapper over the same cells this script grades directly -- the
+# drift guard IS the selftest cell that caught the stale render), and the
+# total is printed. Budget: <45s measured, 60s per leg worst case.
+#
+# rc 0 = docs/TOOLS.md is the render of this tree at HEAD and in the WORKTREE,
+#        tool_map's 20-cell selftest is 20/20, and the repo gate's 8 graded
+#        fast legs are green.
+# rc 1 = a leg graded red.  rc 2 = the repo does not resolve.  rc 124 = a leg
+#        outlived its timeout (a hang is not a pass).
+set -u
+REPO="$(cd "$(dirname "$0")/../.." 2>/dev/null && pwd)" || exit 2
+cd "$REPO" || exit 2
+T0=$(date +%s)
+OUT="${TMPDIR:-/tmp}/toolsmd-drift.$$"
+
+fail() { tail -16 "$OUT" 2>/dev/null; rm -f "$OUT"; echo "TOOLSMD FAIL: $1" >&2; exit 1; }
+
+echo "== 1. tool_map drift guard (HEAD) =="
+timeout 120 python3 tools/tool_map.py --repo . --rev HEAD --check docs/TOOLS.md \
+    >"$OUT" 2>&1 || fail "drift guard at HEAD"
+[ -s "$OUT" ] && tail -2 "$OUT" || echo "   CURRENT: document matches the render of HEAD"
+rm -f "$OUT"
+
+echo "== 2. tool_map drift guard (WORKTREE) =="
+timeout 120 python3 tools/tool_map.py --repo . --rev WORKTREE --check docs/TOOLS.md \
+    >"$OUT" 2>&1 || fail "drift guard in the WORKTREE"
+[ -s "$OUT" ] && tail -2 "$OUT" || echo "   CURRENT: document matches the render of the WORKTREE"
+rm -f "$OUT"
+
+echo "== 3. tool_map --selftest (20 cells, incl. the currency cell) =="
+timeout 180 python3 tools/tool_map.py --repo . --rev HEAD --selftest >"$OUT" 2>&1 \
+    || fail "tool_map selftest graded red"
+grep -q "selftest: 20 cells, 20 passed, 0 failed" "$OUT" \
+    || fail "selftest did not read 20/20"
+grep -qE "^  ok +HEAD: .*the drift guard" "$OUT" \
+    || fail "the currency cell is not among the passes"
+grep "selftest: " "$OUT"; rm -f "$OUT"
+
+echo "== 4. repo gate, 8 graded fast legs (sweep skipped by design) =="
+timeout 180 python3 tools/verify.py --repo . --fast >"$OUT" 2>&1 \
+    || fail "a graded gate leg ran red"
+grep -q "gate: OK" "$OUT" || fail "gate did not read OK"
+grep -q "verify: 9 legs, 8 pass, 0 fault, 0 ungradable, 1 skipped" "$OUT" \
+    || fail "leg tally is not 8 pass / 0 fault / 1 known skip"
+grep -E "verify: |gate: " "$OUT"; rm -f "$OUT"
+
+echo "TOOLSMD OK: drift 0 at HEAD and WORKTREE; selftest 20/20; gate 8/8 fast legs ($(( $(date +%s) - T0 ))s)"