Clean and enhance evaluators that work an anndata

[mengerj]

Description of feature

Evaluation approaches:

• how well do embeddings cluster, based on biological information or batch?
• compared to raw data and or initial embeddings
• scib evaluator
  Annotation quality:
• On a new dataset, if both data and context are embedded, how often is the annotation correct?

[mengerj]

Create a workflow function that takes test datasets and model name as input and applies all relevant evaluation functions to all datasets