Implement classifier aggregating detector results with confidence scoring

[MaximilianRau04]

Context

src/algorithm_pattern_classifier/classifiers/ is currently empty. Once multiple BaseDetector implementations exist (#2, #3, ...), something needs to combine their outputs into a single, ranked answer — a piece of code can plausibly match more than one pattern (e.g. a DP solution that also looks like a nested loop).

What

Implement a BaseClassifier that:

1. Runs all registered detectors against the parsed source.
2. Aggregates the individual detector outputs (which may each independently claim a match) into a ranked list of ClassificationResults with a combined confidence score.
3. Handles ties/overlaps sensibly (e.g. a DP solution shouldn't also be reported as a generic "nested loop" pattern if a more specific detector matched).

Depends on #1, and benefits from having at least two detectors (#2, #3) implemented to test aggregation against.

Acceptance Criteria

• Classifier returns results ordered by descending confidence
• Ambiguous cases (multiple plausible patterns) surface all of them rather than silently picking one
• Unit tests cover: single clear match, multiple detectors firing on the same snippet, and no detector matching