More robust hepevt size please

[jmonk]

I've just had quite a frustrating afternoon hunting down differences in the hepevt common block between my installation of fortran Herwig, HepMC and AGILe and so I feel the need to submit a bug to flag the issue. I realise it is probably impossible to solve because the hepevt size has to be known at compile time.

It turned out that AGILe's call to HEPEVT_Wrapper::set_max_number_entries(10000) in FHerwig.cc was overriding the settings I changed in HepMC to be compatible with the non-Genser version of Herwig that I wanted to use to be compatible with the non-Genser version of Jimmy. Goodness knows what happens to the person unfortunate enough to have unusual length ints, longs or floats (as I mistakenly thought might be the case here on a 64bit kernel).