statistics: add dummyvar for one-hot encoding of numeric and categorical data

[Pasta-coder]

This pull request introduces a new dummyvar function to the GNU Octave statistics package, providing MATLAB-compatible one-hot encoding for numeric and categorical grouping variables.

The implementation:

• Supports numeric vectors and categorical arrays, with column order defined by categories(g) for categorical inputs.
• Treats categorical values consistently with numeric NaN, producing rows of zeros rather than NaN, ensuring safe use in downstream numeric workflows.
• Uses vectorized assignment with sub2ind for efficiency and clarity.
• Includes explicit single-column validation for categorical and table inputs, following Octave API conventions.
• Adheres to Octave style guidelines for argument checking, error messaging, and BIST coverage.