Allow /** */ block doc comments and add bundled-config integration tests

[0xLeif]

Summary

• Config: disable UseTripleSlashForDocumentationComments so swift-format no longer rewrites /** ... */ block doc comments to ///. 0xLeif style uses /// for short (1-2 line) docs and /** */ for longer (3+ line) docs.
• Config: enable lineBreakBeforeEachArgument so partially-split argument lists become one-arg-per-line.
• Tests: add 98 integration tests (35 format + 33 lint + 7 disabled-rule negatives + 1 stress + helpers) that pipe horrible Swift code through the real swift-format binary using the bundled .swift-format.json and verify the exact behavior shipped to plugin users.
• Sources: reformat Plugin.swift and ProtocolHandler.swift with the bundled config so the project dogfoods its own style.

Test coverage

Format mode — tabs/2-space → 4-space, Allman → K&R, blank-line collapse, partially-split args → one-per-line, import sort + @testable grouping, multi-line trailing commas, semicolon splitting, empty closure parens, var-per-line, fallthrough merge, hybrid doc-comment style, idempotence, mixed indentation, trailing whitespace, switch-case label alignment, conditional-compilation indent, range-operator spacing, EOL comment spacing, single-line getter shorthand, OneCasePerLine splits (raw + associated), generic signature wrap, multi-line string preservation, explicit-return preservation, binary operator/colon/comma/paren spacing, trailing-closure call-site cleanup, capture lists, async let, willSet/didSet, property wrappers, some/any, @resultBuilder, #available, single-shot stress test.

Lint mode — force unwrap, force try, force cast, IUO, shorthand types, block comments (with /** */ negative), naming, access-on-extension, parens around conditions, void return, ordered imports, semicolons, forEach, GroupNumericLiterals (decimal + hex), UseSingleLinePropertyGetter, DontRepeatTypeInStaticProperties, UseSynthesizedInitializer, UseLetInEveryBoundCaseVariable, UseExplicitNilCheckInConditions, ReturnVoidInsteadOfEmptyTuple, FullyIndirectEnum, OneCasePerLine, IdentifiersMustBeASCII, NoAssignmentInExpressions, AvoidRetroactiveConformances, AmbiguousTrailingClosureOverload, NoPlaygroundLiterals, NoLabelsInCasePatterns.

Negative cases — confirms NoLeadingUnderscores, OmitExplicitReturns, UseEarlyExits, UseWhereClausesInForLoops, AllPublicDeclarationsHaveDocumentation, ValidateDocumentationComments, BeginDocumentationCommentWithOneLineSummary all stay quiet on input that would otherwise trigger them.

Impact across 0xLeif repos (block-doc-comment rule disabled)

Repo	Before	After
AppState	75	3
Waiter	31	0
Later	66	62
Fork	43	43
Cache	721	627

Test plan

• swift build passes
• swift test passes (98 tests across 5 suites)

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request updates the Swift formatting configuration by enabling line breaks before each argument and disabling triple-slash documentation comments. It also marks the end-to-end verification of the Fledge protocol mode as complete in the task documentation for both formatting and linting. Feedback indicates that the change to argument line breaks is undocumented and potentially significant, and it raises concerns about marking protocol verification as complete while automated integration tests remain outstanding.

[gemini-code-assist]

The change to lineBreakBeforeEachArgument is not documented in the PR summary. This setting significantly alters how function arguments are formatted across the project by forcing each argument to a new line when wrapping. If this change was unintentional, it should be reverted; otherwise, it should be explicitly mentioned in the description to ensure all stakeholders are aware of the style change.

[gemini-code-assist]

The task 'Verify Fledge protocol mode works end-to-end' is marked as complete, yet the 'Add integration test' task remains open. Relying on manual verification without automated integration tests for the protocol mode increases the risk of future regressions.

[gemini-code-assist]

The task 'Verify Fledge protocol mode works end-to-end' is marked as complete, yet the 'Add integration test' task remains open. Relying on manual verification without automated integration tests for the protocol mode increases the risk of future regressions.