docs(#532): record the upstream filing for patch 0020

[gsdali]

What & why

Housekeeping for #532: patch 0020 (BRepFeat_MakeCylindricalHole tool part selection) has been
carried since PR #587 but was never filed upstream. #657's audit (PR #737) confirmed it was ready
to send and drafted the text; this PR does the actual filing and records the result.

Re-verified the defect directly against current upstream master, not just the V8_0_0_p1
baseline the original measurement used: read all four affected call sites and the two correct
callers (BRepFeat_Form, BRepFeat_RibSlot) on master, then compiled
BRepFeat_MakeCylindricalHole.cxx from master as an override translation unit, once unmodified
and once with patch 0020 applied, and ran the existing
Scripts/repro/532-cylindrical-hole-part-selection/
probe against both, linked ahead of the pinned OCCT 8.0.1 archive. The unmodified build reproduces
every figure the existing writeup records as "before", including the oversized-radius
InvalidPlacement status; the patched build reproduces every "after" figure. The touched file is
byte-identical between master and our pin, so this is the same change on both.

Filed as Open-Cascade-SAS/OCCT#1447: a fix PR
with no companion repro issue, per
okf/policies/upstream-occt-style.md and the precedent of
0018, 0019, 0021 and 0024 (an OCCT maintainer's own comment on
OCCT#1409 asked
contributors not to open a separate issue when a fix is already in hand).

This PR touches only CLAUDE.md and Scripts/patches/README.md. No source, test, or public API
change; docs/CHANGELOG.md and docs/SEMVER.md are untouched per
okf/policies/changelog-on-merge.md and
okf/policies/semver-at-release.md (entries below instead).

Not closing #532: the issue itself was closed by PR #587 when the fix was carried; this is a
follow-up on the separate, still-open upstream-filing housekeeping in #657.

CHANGELOG entry

Carried patch 0020 (#532, BRepFeat_MakeCylindricalHole tool part selection) is now filed upstream as OCCT#1447

Re-verified the defect directly against current upstream master before filing: the four affected
call sites, the two correct callers in the same builder, and the second, unfixed defect in the same
heuristic are all unchanged from the V8_0_0_p1-based measurement this patch was originally built
against, and the touched file is byte-identical between master and our pin. Confirmed by compiling
BRepFeat_MakeCylindricalHole.cxx from master as an override translation unit, once unmodified
and once patched, and running the existing reproducer against both: every "before" and "after"
figure in the original writeup reproduces exactly, including the oversized-radius status change.
Filed as Open-Cascade-SAS/OCCT#1447, PR only
per okf/policies/upstream-occt-style.md. No change to OCCTSwift's own behavior or public API.

SemVer impact

NONE. Documentation only (CLAUDE.md, Scripts/patches/README.md); no source, test, or public API
change. The upstream PR proposes a change to OCCT itself, not to this package.

Checklist

• This is a documentation-only change; no new production code path exists to unit-test.
• N/A: no new test or --self-test case in this PR.
• The CHANGELOG entry above is complete, and docs/CHANGELOG.md is not in this diff.
• The SemVer impact above is stated, and docs/SEMVER.md is not in this diff.

Notes for the reviewer

• All five gate scripts and their --self-tests pass locally (check-bridge-index 18/18,
  check-null-handle-guards 24/24, check-docs-defaults 13/13, derive-bridge-header-split --verify 8/8, count-operations consistent at 4306), as expected for a docs-only change.
• The upstream PR title was edited once after opening, to name what it replaces ("not the cut
  result") rather than only what it does, and its body gained a short footnote explaining the
  PerformBlind(20) figure (1178.0972, measured from the axis origin rather than the plate's
  entry face), matching a clarification already present in Ten upstream OCCT PRs now sit on a pre-8.0.1 base, and patch 0020 was never filed at all #657's drafted text.
• Not merging this PR; opened for review only, per the task that produced it.

[gsdali]

The deviation was correct and my dispatch instruction was wrong. Recording that here so the reasoning
is not re-litigated.

I asked for an issue and a PR, "following the repro-then-fix pairing our other filings use", while in
the same prompt telling the agent to read okf/policies/ for upstream contribution style. Those
contradict each other. okf/policies/upstream-occt-style.md says go straight to a PR, quoting OCCT
maintainer dpasukhi on OCCT#1409:

Thank you for your contribution. In case if you preparing PR, no needs to create Issue.

My sentence described the practice that policy exists to retire. Verified independently: the policy
says it, dpasukhi said it, and OCCT#1420 is already PR-only, filed the day after that comment.
Confirmed by the user: PR-only is the policy.

The right outcome, and worth naming because the default is the other way. An instruction in a prompt
reads as more specific and more current than a file, so the usual failure is silent compliance, which
here would have put an unwanted issue on a third-party repo. I have recorded the general rule for
myself: point a dispatched agent at the policy, never paraphrase the process it owns, because a
restated rule is a copy with no update path.

The re-verification is the strong part

Not a text diff against master, which would only have shown the file unchanged. The agent fetched
BRepFeat_MakeCylindricalHole.cxx from current master, compiled it as an override translation
unit, linked it ahead of the pinned archive, and ran Scripts/repro/532-cylindrical-hole-part-selection/
against it. Unpatched reproduced every "before" figure exactly, including the oversized-radius
InvalidPlacement; patched reproduced every "after" figure exactly.

So "still live on master" is measured against a running binary rather than inferred from a hash, which
matters because the whole point of the check was that upstream might have fixed it while we were not
looking.

OCCT#1447 is 4 lines across one file, four
SetOperation(Fuse, true) calls with a one-line comment each, matching the concise style the same
policy requires. That closes the last carried patch with no upstream presence: all fifteen are now
either shipped, filed, or under review.