Postfit improvement

Created by: nhartland

I think what's causing Tommaso's error in NNPDF/nnpdfcpp#65 is that some of his replica files are weirdly corrupted (they cut out after writing out the x-point grid). While that's a problem in itself, we should have a sanity check in postfit(2) to at least ensure that all the replica .dat files have the same number of lines in them (this should be identical replica to replica).

I'm happy to have a pass at postfit2 to add this (and some other checks). So I can self-assign this one. Would everyone be happy with stopping support for postfit1?