Add CLI entry point and labeled evaluation corpus

[MaximilianRau04]

Context

There is currently no way to actually invoke algorithm-pattern-classifier from the command line, and no labeled dataset to measure detector/classifier accuracy against as new patterns are added.

What

1. Add a CLI entry point (e.g. pattern-classifier classify <file>) that runs the classifier pipeline and prints the detected pattern(s) with confidence and supporting evidence (--json for machine-readable output).
2. Build a small labeled corpus of algorithm implementations per pattern (two-pointer, sliding-window, DP, ...) to use as a regression/evaluation set for precision & recall as detectors are added.

Acceptance Criteria

• pattern-classifier classify path/to/file.py runs end-to-end and prints a report
• Console entry point registered in pyproject.toml
• Labeled corpus lives under tests/ (or a dedicated benchmarks/ fixtures directory) with multiple examples per pattern
• CI runs the classifier against the labeled corpus and reports precision/recall (or fails on regression)