enhancement request (qol improvement): Append temporary tags via datasets TOML

Lower priority enhancement request.

Some models (Such as WD1.5) benefit from the addition of model-specific tags for realistic vs anime stylized datasets. Being able append these in bulk without directly modifying the tags/captions files (e.g. by specifying additional tags to add to a particular dataset in the toml file) would make managing the swapping of datasets for training between various base models easier.

No files or tests are named. Start by locating the dataset TOML parsing and tag/caption loading entry points, then determine how a dataset-specific list of additional tags can be applied without modifying source files; done means those tags can be configured per dataset and persist when swapping datasets for different base models.

Written by the indexing model from the issue text.