docs: transcribe the 13 outstanding CHANGELOG entries, and stop miscounting one

[gsdali]

Release step 2 is "check the ## Unreleased section is complete". It was not.

Scripts/check-changelog-transcription.py --verify-transcribed reported 13 merged PRs whose
## CHANGELOG entry was written in the PR body and never copied into docs/CHANGELOG.md. That is
my omission as merger, not the authors': changelog-on-merge
puts transcription at merge time, and I merged these without doing it.

Transcribed here in merge order, newest first: #800, #804, #798, #797, #799, #754, #733, #741,
#740, #738, #730, #743, #737. (#741 contributed three entries.)

Two needed a heading synthesised, because their PR body gave the entry another shape:

• fix(#636): guard GeomAPI_ExtremaCurveCurve against parallel curves #730 wrote prose with no heading. Given ### \Curve3D.extrema(with:)` no longer crashes the
  process on parallel curves (Curve3D.extrema SIGSEGVs on parallel curves at every capacity, including its default #636)`.
• chore(#512): pin v2.0.0-kernel.2, the kernel holding all fourteen carried patches #754 wrapped a bullet in a ```markdown fence. Given
  ### Kernel pin moved to \v2.0.0-kernel.2`, carrying all fourteen patches (Rebuild OCCT.xcframework with patch 0017 (#484 null-context guard) and ship a patch release #512)`.

One detector fix

DECLARED_NONE gains a #-prefix allowance. #751 answered "no entry needed" as a heading
(### None, investigation only, no functional change (#597)), and the anchored regex read that as a
real entry naming a symbol called None. Nothing could ever transcribe it, so it would have reported
MISSING forever, in a report a human is meant to trust at release.

One new self-test row. Proven load-bearing: reverting the allowance fails it and two neighbours
(23/23 -> 20/23).

Result

transcribed late, nothing to do:              23
entry in the PR body, NOT in the CHANGELOG:    0
PR body says None, by design:                 13
PR body has no entry section at all:           1   (#757, already noted in the file)
could not verify:                              0

The file's own TRANSCRIPTION NOTE is rewritten to match; it still described the earlier state (ten
transcribed, six declared-none, "expect it to keep reporting 31").

The bare run still reports 37 and always will. It asks whether the merge commit itself touched
the file, which is the right question at merge time and unanswerable afterwards without rewriting
history. --verify-transcribed is the one to read for whether anything is genuinely missing.

Why this is a PR and not a push

changelog-on-merge.md has the merger commit the transcription onto the base. The repository
ruleset makes gate-scripts a required check on refactor/381-pass1b, which declines a direct
push. Worth reconciling in the policy after the release; noted, not changed here.

CHANGELOG entry

None. This PR is the transcription; an entry describing it would be circular.

SemVer impact

NONE. Documentation and one gate-script detector fix. No public API is touched.