Work out weighted cross-section average in multi-file runs

As reported by @fsiegert - when running Rivet over multiple HepMC files in the same run, the cross-section used in finalize will correspond to the one reported by the last file.

This change set is to work out an on-the-fly weighted average (using the effective number of entries as the file weight) when running over multiple HepMC file and use that instead.

(Doesn't affect merging but is relevant for reentrant/"top up" runs.)