feat: Add inline suppression comments for complexity thresholds

[fummicc1]

Summary

Adds // swift-complexity:disable so a single, reviewed function can be exempted from threshold checks without raising the threshold globally or abandoning enforcement. This closes a real gap left by the per-type threshold work (#32): today, one legitimately complex function forces a choice between fixing it or weakening the check for the whole type.

Second step of the "Now" roadmap, after SARIF output (#33).

Syntax

// swift-complexity:disable
func parseLegacyFormat(_ input: String) -> Document { ... }

// swift-complexity:disable cognitive
func stateMachine(_ event: Event) { ... }  // cyclomatic is still checked

• The comment must be the closest leading comment directly above the declaration it suppresses (function, initializer, deinitializer, or computed property/accessor).
• A bare disable suppresses every metric; naming specific metrics (cyclomatic/cognitive, space- or comma-separated) suppresses only those.
• Deliberately no "next line" form and no "enable" counterpart — suppression always scopes to exactly one declaration, so unlike range-based disable/enable there's nothing to forget to re-enable.
• Fails closed on a typo: an unrecognized metric name suppresses nothing for that comment, rather than falling back to "suppress everything." A typo should not silently widen what a complexity tool hides.
• /// doc comments are never mistaken for a directive.

Design notes

• Suppressed values are still computed and shown in every output format — only the threshold judgment is skipped. This meant updating all three places that independently decide "does this function violate its threshold" so a suppressed metric can't leak through one of them: ThresholdConfiguration.isExceeded (feeds the exit code and the upstream filter shared by text/json/xml/xcode/sarif), the Xcode diagnostic (which reports both metrics in one combined message and keeps doing so — only the gating changes), and SARIF's per-metric result generation (which computes its own >= check independent of isExceeded).
• LCOM4 class cohesion has no suppression yet — it's a type-level metric, not a per-function one, so it needs its own design. Left as a follow-up.
• New --report-suppressions flag prints every suppressed function and its current metric values to stderr, so suppressions stay visible instead of silently accumulating.

Verification

• swift test: 102 tests pass, including new coverage for SuppressionParser (bare disable, specific metrics, comma-separated, typo'd metric, unrelated comment, doc comment, false-prefix match like disableFoo), ThresholdConfiguration.isExceeded with suppression, Xcode diagnostics with partial/full suppression, and SARIF with partial suppression
• E2E against a real fixture (Tests/SwiftComplexityCoreTests/Fixtures/suppressed_functions.swift) confirmed with the built CLI across text, xcode, and sarif formats at multiple thresholds, and --report-suppressions output
• swift-format clean via lefthook pre-commit; no new markdownlint errors introduced (verified by diffing lint output against main)

https://claude.ai/code/session_01GaTHVDhSAx76RbLmVziLQu

[github-actions]

📊 PR Analysis

Type	Count
Swift files	12
Test files	4
Documentation	2
Configuration	0

📝 Large PR: Consider breaking this into smaller, focused PRs.

---

Automated analysis by swift-complexity CI

[github-actions]

🔍 Complexity Analysis Results

✅ All functions are below the complexity threshold of 15.

This analysis was performed by swift-complexity itself.

[github-actions]

📊 PR Analysis

Type	Count
Swift files	12
Test files	4
Documentation	2
Configuration	0

📝 Large PR: Consider breaking this into smaller, focused PRs.

---

Automated analysis by swift-complexity CI

[github-actions]

🔍 Complexity Analysis Results

✅ All functions are below the complexity threshold of 15.

This analysis was performed by swift-complexity itself.