Feat: add CO2CONTENT.m algo,

[RaphaelBajon]

This pull request introduces a major new feature to the canyonbpy package: the addition of the CONTENT method for generating internally-consistent carbonate system estimates, alongside significant improvements to input broadcasting and documentation. The update also refines the handling of input shapes and coordinates, making predictions more robust and user-friendly when working with multi-dimensional xarray datasets. The package version is bumped to 0.4.0 to reflect these changes.

New CONTENT functionality:

• Added the content method to the CanyonBAccessor class, enabling users to compute internally-consistent carbonate system variables (AT, CT, pH, pCO₂) by combining CANYON-B neural network outputs with CO2SYS estimates, including uncertainty propagation and options to include raw or full sub-outputs.
• Implemented the _pack_content_results method to correctly package CONTENT outputs as xarray.Dataset objects, preserving dimensions and coordinates, and supporting optional output details.

Input broadcasting and shape handling:

• Improved the DatasetToNumpy class to broadcast all input variables to a common shape before flattening, ensuring compatibility with multi-dimensional datasets and robust reshaping of outputs. Added original_dims to track dimension names (following feat: native xarray support via ds.canyonb accessor (v0.3.0) #7).
• Updated result-packing logic in both _pack_results and _pack_content_results to use the broadcasted shapes and dimension names, ensuring that output datasets align correctly with their input grids (following feat: native xarray support via ds.canyonb accessor (v0.3.0) #7).

Documentation and usability:

• Expanded the accessor docstrings to document the new CONTENT workflow, provide clearer usage examples, and clarify input requirements and options.
• Improved parameter documentation for input error arguments in canyonb and clarified default values and units.

Tests:

• Updated tests to import and check the new CONTENT functionality and ensure merge compatibility with xarray.