Support schemas

[nicolay-r]

UPD
I think it is even better option is to merge here
https://github.com/nicolay-r/bulk-ner/blob/main/bulk_ner%2Fapi.py#L48-L49
Because pipeline should not know about various inputs. So that, we just just stacking dicts and attempting to merge them with no intersections.

And this is where we refactoring horizontal batching:
https://github.com/nicolay-r/bulk-ner/blob/main/bulk_ner%2Fapi.py#L37

Old reasoning
I think we have to add grouping of the rows concept in the pipeline here
https://github.com/nicolay-r/bulk-ner/blob/main/bulk_ner%2Fapi.py#L23-L32
As the list item.
At present, supporting schemas would lead to amount of rows in output similar to n*c, where c is the total amount of columns