Record logic for some processors may be wrong

[hunterhector]

Describe the bug
For example, in the Spacy record function:
https://github.com/asyml/forte-wrappers/blob/main/src/spacy/forte/spacy/spacy_processors.py#L280

The record_meta for the Sentence type is first set to an empty set. However, this may not be true if there are existing processors that already add Sentences.

The problem is that what should we do when two processors both add the Sentence type? We probably should maintain separate records for each processor, something like:

{
    "spacy": "ft.onto.base_ontology.Token": {"ner", "pos"}
    "stanza": "ft.onto.base_ontology.Token": {"lemma", "pos"}
}
``