Catching case where first truth vertex is nullptr

There are cases where the first truth vertex in a record is a nullptr, but subsequent truth vertices are fine. This protects the code against such situations.

This is a weird one. Some generators don't assign signal process vertices, which is what we're really looking for here, so in newer releases they are set like this: https://acode-browser1.usatlas.bnl.gov/lxr/source/athena/Generators/EvgenProdTools/src/FixHepMC.cxx#0038

which I believe is equivalent to this code. It seems that the 0th vertex is being dropped during reconstruction, which is also weird, but it is there during evgen. I just opened AGENE-1803 for this, so that it can be tracked there.

Thanks to @jonesel for the report.