Modernise InputCopyStream

Whilst the nominal deployment of HLT2 at the pit does not require it, there are other use cases (development, MC productions) that need Moore to directly write out a data file.

Quoting @apearce

"Having a replacement for the InputCopyStream algorithm would be nice (it doesn’t work with the new scheduler as it relies on incidents). The current logic in HLT2 doesn’t propagate any information from the input file, including MC information which is needed for physics studies. Because we happen to know this location we can hard-code the copying, but I’d like something more dynamic and generic. The algorithm logic won’t be complicated, the trick will be figuring out how to collect the list of locations to be copied, which might mean digging into Gaudi internals (and/or asking Marco Clemencic or Markus)."