Updates to DNF/CNF methods

[daemontus]

This PR fixed two problems:

• The output of the DNF optimization algorithm was in some cases less optimized than we hoped for. This new approach should perform better in case the BDD contains a lot of shared sub-formulas (which would otherwise contain a lot of extra literals and sometimes clauses).
• The DNF/CNF construction had a bug due to which it was dependent not on the size of the DNF/CNF, but on the number of paths in the resulting BDD, which can be sometimes substantial. The new algorithm is a bit more high-level (and should have more overhead), but its complexity shouldn't explode unless the resulting BDD explodes.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 63.59%. Comparing base (23d4c53) to head (75d8572).

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #57 +/- ## =========================================== - Coverage 90.87% 63.59% -27.29% =========================================== Files 27 26 -1 Lines 2093 1420 -673 =========================================== - Hits 1902 903 -999 - Misses 191 517 +326 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.