Make data cleaning optional in the LSA primitive

[thehomebrewnerd]

Currently the LSA primitive implements a cleaning step before computing feature values. This cleaning removes punctuation, lowercases the test, removes stopwords and performs lemmatization, and also adds time to the calculation process. We should make this cleaning step configurable, so users that might not want the cleaning applied can optionally skip this step. To maintain current behavior, the default should be to apply cleaning.