Preprocess Text + Collocations: enable including n-grams

**Is your feature request related to a problem? Please describe.**
It is often desired to include only specific n-grams into the analysis. For example, n-grams based on PMI and some frequency metric. Currently, it is not possible to do this in Preprocess Text.

**Describe the solution you'd like**
Pass a list of selected words (n-grams) from Collocations to Preprocess Text. This could be of type Words. Use Words input for filtering.

**Describe alternatives you've considered**
Save the list with Save Data. However, we would still need to enable using n-grams in filtering (I don't think it currently works). And applying the filter only to n-grams, not 1-grams (or something).

**Additional context**
Words, such as "the European Union", "Jules Verne" or "cutting board" should be included as n-grams, rather than single tokens.

Start by reading the Preprocess Text and Collocations widgets, including their Words inputs and filtering behavior. Trace how selected n-grams could pass from Collocations into Preprocess Text, then verify that multi-word terms are retained as n-grams while the requested filtering scope is respected.

Written by the indexing model from the issue text.