ML-based Anomaly Detection for data content

[vb-dbrks]

Changes

This PR adds ML-based anomaly detection to DQX, enabling users to detect unusual patterns in their data that can't be caught by traditional rule-based checks.

Key features:

• Auto-discovery: Automatically selects relevant columns and creates segmented models when needed
• Isolation Forest: Uses scikit-learn's Isolation Forest algorithm for fast, scalable anomaly detection
• Explainability: SHAP-based feature contributions show why records were flagged
• Unity Catalog integration: Models stored in UC with full lineage and versioning
• New check function: has_no_anomalies() works like other DQX checks
• Production defaults: Ensemble models (2x), 0.60 threshold, contributions enabled by default

What's included:

• New AnomalyEngine for training models
• Feature engineering for numeric, categorical, datetime, and boolean columns
• Model registry with drift detection
• demo 101
• documentation updates

Resolves #957

Tests

• manually tested (ran all demos on Databricks)
• added unit tests (124 tests across 7 test files)
• added integration tests (200+ tests covering training, scoring, ensemble, drift, etc.)
• added end-to-end tests
• added performance tests