better identification of elastic networks

[csbrasnett]

at the moment the elastic network has to be specified by -ef, which is annoying if you can't remember what you exactly used, or if you have some kind of optimised network that doesn't have the same strength everywhere (this is also true if you've used -ea or -ep in martinizing the protein in the first place).

This could be fixed by some kind of analysis of BB bonds which aren't direct graph connections, i.e. don't have adjacent resids