[cmccomb]

Scripts can now choose regularization strength with train(x, y, "lasso", #{ alpha: 0.1 }) or the corresponding logistic-regression call. Existing three-argument calls and empty options maps retain the previous defaults. Unknown options, unsupported linear-regression options, negative or non-finite values, and wrong types return Rhai errors.

This prepares 0.1.4 with alpha as the only new option. It adds a held-out comparison example and keeps the README at 50 lines; detailed usage stays in the API reference and usage guide.

Validation:

• 29 behavioral tests pass, including 12 new compatibility, validation, and parameter-effect tests.
• Three generated Rhai examples and three Rust doctests pass with metadata enabled.
• Formatting, strict Clippy, documentation, and fresh packaged-consumer checks pass locally.
• CI runs the suite and comparison example on Linux, macOS, Windows, and minimum Rhai 1.16.2.