use `pie-core`

[ArneBinder]

The idea is to clean up this repo by outsourcing the core functionality which will advance in a different speed than the remaining content of this repo which will primarily be:

• pipeline,
• model with PyTorchIEModel and AutoPyTorchIEModel (renamed from AutoModel)
• taskmodules and models from pie-modules
• some utility / helper methods.

See https://github.com/ArneBinder/pie-core for reference. The code in that repo will be much more tested as it is currently the case, the target is a test coverage comparable to code in pie-modules (95%) and pie-datasets (98%).

Note that the responsibility of the remaining content overlaps quite a bit with pie-modules. The future will hopefully bring more insights what the distinct roles of both of these packages will be. Some more context: It is planned to also have packages for feature based implementations (e.g. based on sklearn; there are already implementations available in some other projects) and for LLM based IE setups.

Follow-up:

• Do not re-export all individual modules of pie-core (but this will be breaking). Everything exported from pie_core should still be re-exported (e.g., from pie_core import * at the project root), so that from pytorch_ie import Annotation still works, but not pytorch_ie.core.document import Annotation
• Remove AutoPipeline (also breaking).