feat(pdftract-2iyk): implement classifier engine

Implements Phase 5.6.2 classifier engine that evaluates document type
profiles against extracted feature signals.

- ClassifierEngine: evaluates profiles, computes normalized scores,
  returns highest-scoring profile above threshold
- FeatureSignals: struct containing all metrics for predicate matching
- ClassificationResult: document_type, confidence, reasons, runner_up
- Score normalization: matched_weight / total_weight to [0, 1]
- Predicate evaluation: all MatchPredicate variants supported
- Regex caching: OnceLock-based cache for TextMatchesRegex
- Unit tests: 28 tests covering invoice, scientific_paper, unknown
  classification, score normalization, tie-breaking, determinism

Closes: pdftract-2iyk

crates/pdftract-core/src/profiles/mod.rs

@@ -17,9 +17,13 @@
 //! are the shared vocabulary between the rule engine, built-in profile definitions,
 //! and user-authored YAML profiles.
 
+mod engine;
 mod loader;
 mod types;
 
+pub use engine::{
+    classify, has_currency_pattern, ClassificationResult, ClassifierEngine, FeatureSignals,
+};
 pub use loader::{check_forbidden_keys, ForbiddenKeyError, ProfileLoadError};
 pub use types::{MatchPredicate, Profile, ProfileType};