Changed
Types are now generic in dtypes as well as shapes. For example, instead of `Array2[Batch, Time]`, one can now write `Array2[uint8, Batch, Time]`.
Note that this is a breaking change. If you want to retain the old behaviour, you can import `ArrayNAnyDType` instead of `ArrayN` - eg:
python
from tensor_annotations.jax import Array2AnyDType as Array2