Add mechanical checks from plain-writing guides

[hamelsmu]

• en-dash: catches the Unicode dash in ranges such as 10–20.
• ai-vocabulary: catches stock LLM words such as delve, robust, leverage, and unlock.
• rhetorical-question-chain: catches two or more consecutive questions.
• vague-attribution: catches unsourced phrases such as experts say and studies show.
• count-opening: catches openings such as three things and two reasons.

[Copilot]

Pull request overview

This PR adds six new mechanical prose-lint patterns (all model-free) and wires them into configuration and documentation, expanding the existing set of slop/pattern checks.

Changes:

• Added new pattern detectors and Pattern definitions for en-dash, middle dot, stock AI vocabulary, consecutive question chains, vague attribution, and count-first openings.
• Added config flags (TOML + defaults + merge/apply) to enable/disable each new rule.
• Updated README example configuration to document the new pattern keys.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/patterns.rs	Implements six new detectors, registers them in active_patterns, and adds unit tests for the new detectors.
src/config.rs	Adds new pattern flags to the resolved config, TOML shadow types, defaults, and merge logic.
README.md	Documents the new [patterns] keys and brief descriptions in the sample config.

Suppressed comments (1)

src/patterns.rs:819

• The new patterns add several config flags and active_patterns wiring, but the tests here only cover the individual detectors. Adding a unit test that verifies the new config flags actually disable the corresponding patterns would prevent regressions in the active_patterns wiring.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.