Revised domain detection and feature preparation pipeline

[segef]

The domain detection pipeline is revised as follows:

• domains are detected iteratively, one iteration per each domain
• relaxed the domain detection thresholds (tmscore to 0.2, and num of aligned residues to 50-70)
• clean-up of the unused methods and deprecated files
• introduced a new script for the calculation of the structural features, which will be used by the classifiers in later stages (final pkl structure is the same as before to keep compatibility with the classifiers)
• new notebook for clustering domains
• support for config files in domain detection pipeline
• optional prefiltering of sequences by foldseek alignments to domain templates (if no alignment detected, sequence will be ignored)

structural data