Accept optional metadata at the feature extraction step

[rth]

It might be useful to save additional metadata associated with each document in the collection when performing feature extraction.

For instance, in the case of email documents, it could be the To, CC, Subject etc fields, that would be used email threading.
This would allow in particular to remove the /api/v0/email-parser/ endpoint altogether.