v1.3.2: Modernization stage 2

[nitinmad]

Summary

This PR continues ChemML modernization, with AutoML updates to improve usability, bringing Keras 3 compatibility, and migrating CI/CD infrastructure from Travis CI to GitHub Actions, with cross-platform testing on Ubuntu, macOS, and Windows.

Key Changes

AutoML Enhancements

• Enhanced AutoML multi-core processing with user selection for running only lighter models; much faster runtime, but potentially less accurate models as a result
• Added XGBoost to regression model space for AutoML
• Added Mordred descriptors as featurization option for AutoML
• GradientBoostingRegressor and SVR now unavailable in AutoML for datasets>1000 rows due to exponential computational resource increase

Keras 3 Compatibility

• Removed unsupported decay parameter from Adam initialization
• Fixed multi-output model metrics using dictionary-based mapping to prevent duplicate metric naming
• Added mixed precision policy initialization to prevent ml_dtypes.float4_e2m1fn compatibility errors with TensorFlow 2.19+
• Updated LorentzLorenz model compilation for Keras 3 requirements

CI/CD Infrastructure

• Created GitHub Actions workflow with cross-platform matrix (Ubuntu, macOS, Windows) × Python 3.12
• Replaced Travis CI with native GitHub Actions (removed .travis.yml)
• Removed Codecov dependency in favor of GitHub-native artifact uploads for coverage reports

Other Features and Improvements

• Added get_params() method to MLP class for ChemML AutoML compatibility
• Updated README with GitHub Actions badge and GitHub releases link
• Comprehensive CHANGELOG documenting v1.3.2 release

Bug Fixes

• Fixed unreachable else statement in mlp.py _parse_opt_config method
• Fixed openbabel import failures in CI/CD by adding openbabel-wheel
• Removed unused turtle import from explain.py that was causing macOS CI failures

[nitinmad]

PR tests passed. ignoring codecov as it is outdated.