Clarify brain and evaluation keys

[ehofesmann]

Stages that require knowledge of evaluation and brain keys often end up using exactly what is in an example rather than the provided keys from the dataset.

For example the query show me false positive model predictions with a high confidence returns the view:

• filter_labels("det_model",(F("eval_det") == "fp") & (F("confidence") > 0.8))

Where eval is the field on the dataset but eval_det was used in an example.

For brain runs, sort_by_similarity stages also end up using the exact brain_key from an example.