Refactor pipeline interface

[mrchtr]

Feature branch for refactoring the pipeline interface.

Going to create separate PRs to maintain an overview of all changes and avoid breaking the main branch.

[RobbeSneyders]

FYI, the docs pipeline is failing because we reference the Pipeline class in the docs.

[RobbeSneyders]

Thanks @mrchtr, great work!

I haven't reviewed the tests yet (the PR is quite big 😅), but already left my comments on the code.

[RobbeSneyders]

Suggested change

	f"Found multiple instantiated datasets in {module_str}. Use the first dataset to start "
	f"the execution."
	f"Found multiple instantiated datasets in {module_str}. Using the first dataset to start "
	f"the execution."

Shouldn't we use the last dataset? Or should it not matter? And if it doesn't matter, should we even log this?

[mrchtr]

I've checked it again. We don't need it since we don't create a deepcopy of the graph. Every dataset will contain the same computational graph after applying a new operations. We are not able to create a deepcopy since the graph contains some mappingproxy objects. Shouldn't be a problem for now, if we just allow a single dataset object in a module.

For the case we allow something like this:

dataset_1 = Dataset.read(...)
dataset_2 = dataset_1.apply(...)
dataset_3 = dataset_2.apply(...)

we will run into issues. For now we could add a validation that every dataset of a module contains the same graph, wdyt?