…ady included

Add a command-line option to exclude "unknown protein" or "unknown sequence family" from results

Currently, phrases, i.e. sub-sets of candidate descriptions, that only consist of non-informative words are still scored and might be assigned as the final annotation. This should be changed.

If the set of informative words is empty, classify the protein or sequence family as "unknown".

The new option is '-x'

- Replaced standard Rust regex with fancy-regex in capture-replace-pairs, thus
  allowing for (named) backreferences.
- Using named backreferences multiple occurrences of the same word are replaced
  with the first occurrence, i.e. any subsequent occurrence is deleted.

- Polishing iteratively applies capture-replace-pairs (fancy-regex, replace instruction)
  to the human readable descriptions (HRDs) assigned to the queries (families or proteins)
- This is used among others to remove terminal non-informative words like '[...] and', or
  [...] the', or '[...] or' etc.
- The polishing step can be suppressed (skipped) by providing the new command line option
  --polish-capture-replace-pairs (-d) with "none". Use the same option to provide
  custom capture-replace-pairs.

- problem was mutability and iteration over mutable references...

Example polish-capture-replace-pairs file in misc.