Features to add

[benjimaclellan]

• Constrain Dataset type in model definition (i.e., an annotated dtype)
• Pipeline define on keys or Groups? Currently multiple groups of the same type can be in a datastore, distinguished by the key
• How to handle different testbench designs, e.g., different number of oscilloscopes - such that parsing doesn't break
• In Dataset definition, allow specified dimension (i.e., number of axes) as well as None or specified shape (which are currently supported)
• Parsing undefined models, and more error/warning, so that h5 files can be loaded even if model is unknown
• No extras allowed in Pydantic models
• More error handling & messages when defining models/initializing models
• Handling of flexibly sized Groups where the amount of datasets is not fixed.