More IO through unified interface + benchmark improvements

[bittremieux]

• More IO options:
  • Peak files: MGF, mzML, mzXML
  • Clustering results: MaRaCluster, spectra-cluster, MS-Cluster
  • Identifications: mzTab, idXML, MaxQuant (We need to make sure the identifications use scan numbers instead of indexes. This needs to be double checked.)
• Unified interface to representative selection in representative.py.
• Benchmarking script has been cleaned up and now does very fast average dot product computation in evaluate.py.
• Misc:
  • I've omitted the ProteomeXchange ID from the spectrum references for simplicity. Spectrum references take the abbreviated form "{filename}:scan:{scan}" now instead.
  • Filter cluster by minimum number of spectra.