Increasing min_rep doesn't reduce cluster memory usage

[tnitka]

Unsupervised clustering has the same memory usage with values of min_rep that exceed the number of input sequences as with min_rep: 0. This occurs regardless of whether the Blazing Signature Filter or SciPy is used to calculate Jaccard distance, but the error message when using SciPy indicates that Snekmer is running out of memory during this calculation.