Speed up views generation

[benel]

For now, different views have a loop on words (corpus_lexicometrics, document_lexicometrics, kwic, phrase). To improve views generation performance, we could try to emit in the same loop views that have similar keys and reduce functions.

This seems to be the case of phrase and corpus_lexicometrics:

for each word1 {
  get word2 and word3
  if you have word3 {
    emit([word1, word2, word3])
  } else {
    emit(word1)
  }
}

[benel]

Note: This algorithm is just an illustration. There is a more optimized way to do that by getting words only once and remembering the 2 immediate previous words.

[benel]

No gain in generation time with the all-in-one view (c844940) compared to corpus_lexicometrics + document_lexicometrics + phrase (e46c11e). :'(