De-jargon-ify the change-detection work

[imnasnainaec]

Onto docs/terminology-alignment (#22), which branched before
502910d landed on main. Two commits:

1. main merged in (not rebased). One conflict, in
   docs/en/guides/bulk-edit-glosses.md, where De-jargon-ify #22 reworded the
   validation bullet that 502910d inserted new bullets above; resolved by
   keeping both, verbatim. git diff main HEAD on that commit is exactly
   De-jargon-ify #22's own diff, nothing more.
2. The same terminology decisions applied to the change-detection work —
   the Changes / RangesChanges docstrings, the Lexicon query docstrings,
   node_diff / _nodes_aligned, the two new guide bullets, and the new test
   docstrings and comments. Prose only.

Terms #22 had already ruled on

In 502910d	Here
"the passthrough layer declined to scan"	"the byte scanner declined to read"
"the same work save() does to decide passthrough"	"…to decide which source bytes it can reuse"
"the passthrough matrix pairs slot i with node i"	"byte reuse pairs the ith source region with the ith node"
"fill the source's slots one-for-one"	"match the source's regions one-for-one"
"an entry's digest spans its subtree"	"…covers its whole subtree"
"ranges span several files"	"ranges live in several files"
"root-level residue" (bare)	"root-level LIFT residue"

span → region and the loss of "matrix"/"slot" follow c55c8c9; the LIFT-residue
qualifier follows db67388. "passthrough" is left standing where #22 keeps it
(CHANGELOG.md, _canonical.py, _scan.py, the reader/writer test names) —
only the "layer" and "decide passthrough" uses are reworded, matching how #22
treated the same phrases.

New jargon in 502910d

Not covered by #22, since the words weren't in the repo when it was audited:

• mtime — an unexpanded abbreviation, in both the changes() docstring
  and the guide. Spelled out as "file-modification time".
• "membership held", "a permutation alongside one", "a set difference
  would swallow it" — set-theory borrowings, in the Changes /
  RangesChanges docstrings, node_diff, and a test docstring. Now "the same
  entries are still present", "a reordering alongside one", and "comparing the
  two lists as sets would miss it".
• "the composite" for a Changes — a coined sense, in a test comment. Now
  "the whole-document report".
• "the same hole … of the guard" — a test docstring; now "the same gap".
• "the change guard stays honest about them" — _nodes_aligned; now "stays
  in step with the writer", which is the actual claim.

Deliberately left alone: falsy / truthy (Python's own vocabulary),
digest, baseline, and aliased — the last glossed on first use in
node_diff ("appending a node the document already has aliases the object
rather than copying it") and in the _duplicate_entry test helper.

Verified

python scripts/check.py green (ruff, format, mypy --strict, 529 tests at
97.6% coverage); mkdocs build --strict green.

---

This change is