Make Pre-processing options work for PreTrainedVectorizer

[dafajon]

Currently get_pretrained_embeddings, get_bert_embeddings work on the raw form of the document. As a result preprocessing settings do not apply to the text that goes into the transformer based vectorizers.

• Add ignore_preprocess option to vectorizer to use raw text.
• Build input str sequence from filtered Token objects before passing it to the SentenceTransformer.encode method.