Add YAML/JSON configuration manager for hyperparameter management (fixes #208)

[kamalahasiniburra]

Summary

Addresses Issue #208. This PR adds a ConfigManager class that abstracts hyperparameters into YAML/JSON configuration files for reproducibility and agentic automation.

Features

• ConfigManager class with dot-notation access (e.g., config.get('training.learning_rate'))
• • Built-in defaults for classification and regression tasks
• • Load/save JSON and YAML configuration files
• • Deep merge: partial config files are merged with sensible defaults
• • Configuration validation (learning rate bounds, data split consistency, etc.)
• • Configuration diff: compare two configs to see what changed

Tests

• 25 comprehensive tests, all passing
• • Covers: defaults, get/set, file I/O, validation, diff, and dict operations

Files

• config_manager.py: Main configuration manager module (312 lines)
• • test_config_manager.py: Test suite (197 lines)