kundajelab/tfmodisco: Lower Mem AGKM + Variable Len Seqlet ID + Initial Exemplar-Based Hit Scoring

Corresponds to PR https://github.com/kundajelab/tfmodisco/pull/78 AGKM embeddings are now the default (also avoids importing tensorflow unless user wants to use the previous gapped kmer embeddings) cap on the agkm embeddings size; reduces memory use Implementation of variable-length seqlet identification Initial implementation of exemplar-based hit-scorin