chore(#114): trial the ecosystem code-style policy

[gsdali]

What & why

Pilot repo for the ecosystem's proposed code-style policy
(ecosystem docs/code-style-policy-proposal-2026-08.md),
chosen deliberately ahead of OCCTSwift in the rollout order: this codebase is small (48 files,
~10k lines) so it's one we can fully sweep into compliance in one PR and turn on zero-tolerance
blocking CI immediately, rather than needing OCCTSwift's gradual "if you touch it, you fix it"
exemption-manifest approach. Also the first real test of the CI gates themselves.

Adds:

• .swift-format (100 col, 4-space — the ecosystem's one deliberate divergence from Google's
  2-space default) + a full swift-format --in-place sweep of Sources/+Tests/.
• .swiftlint.yml, deliberately narrow (only_rules: [orphaned_doc_comment]). SwiftLint's
  default rule set duplicates swift-format's formatting opinions (can disagree with them on the
  same line) and separately opens a large code-quality/complexity surface
  (identifier_name, cyclomatic_complexity, function_body_length, nesting, ...) that
  overlaps the ecosystem's own code-structure policy rather than this one — identifier_name's
  default alone produced 548 hits on first run, all short conventional locals (i, db, g,
  sv) in dense numerical/graph code, none real problems. Scoped down to the one rule that
  catches something swift-format has no equivalent for.
• Scripts/comment-ratio-check.sh: report-only nudge (never fails), matching this repo's own
  bash-script convention rather than the Python style OCCTSwift/Scripts/ uses.
• .github/workflows/code-style.yml: swift-format lint --strict and swiftlint lint --strict
  as blocking gates, comment-ratio as a non-blocking step. Pure text analysis, no swift build
  needed, so it stays cheap even on the macos-15 runner it uses for tool availability.
• okf/policies/code-style.md: this repo's 8th mandatory policy, linked from okf/index.md and
  CLAUDE.md (count bumped seven → eight).
• 47 by-hand fixes for lint-only findings swift-format --in-place can't auto-fix: doc-comment
  structure (BeginDocumentationCommentWithOneLineSummary), two Throws: additions and one
  Parameters: pluralization (ValidateDocumentationComments), and two snake_case → camelCase
  renames in Sources/Script/main.swift (c_seg1..10/w_seg1..10 → cSeg1..10/wSeg1..10,
  local-only, all call sites updated).

Explicitly out of scope here (see the issue): clang-format (no first-party C++ in this
repo), and the Swift-symbol docs-drift checks proposed for OCCTSwift — this repo's
docs/reference/ documents CLI verbs, not Swift declarations, and Scripts/verb-check.sh /
.github/workflows/verbs.yml already guard that registry-vs-docs consistency.

Zero behavior change. Renames are local-only; everything else is formatting/doc comments.
Verified directly, not just asserted:

• swift-format lint --strict, swiftlint lint --strict, Scripts/comment-ratio-check.sh,
  Scripts/policy-check.sh all pass locally.
• Release build succeeds; Scripts/verb-check.sh and Scripts/run-identity-check.sh both pass
  against it.
• swift test --filter DrawingComposerTests: 3/3 pass (covers the hand-edited files).
• Scripts/recipe-check.sh: all 7 recipes pass, every one with a reference output.brep matches
  it at Δvol = 0.00e+00 — same geometry, byte-for-byte, before and after the sweep.

One pre-existing, unrelated finding surfaced during verification: a plain debug swift build hits
an OCCTSwift/OCCTSwiftAIS sibling-version link error (Face.SurfaceType symbol not found).
Confirmed present on main before this change too (via git stash + rebuild). The release build
CI actually uses (swift build -c release) is unaffected. Not fixed here, out of scope for a
style/docs pass; flagging in case it's worth its own issue.

Closes #114

Checklist

• New or changed behavior is covered by a unit test in the same PR (not just manual
  verification); see SecondMouseAU/OCCTReconstruct#397
  for the ecosystem-wide test-coverage standard this is piloting.
  — N/A here: no behavior changed (formatting/docs/CI only), verified via the existing
  DrawingComposerTests suite and recipe-check.sh's reference-output comparison rather than
  new tests, since there's no new behavior for a new test to cover.

Notes for the reviewer

Per the ecosystem's own gate (no repo implements without a filed issue first), this was filed as
#114 before any of this landed. Holding for review before merge, per "we'll review and implement."

🤖 Generated with Claude Code