Fixes for automatic systematics

Fixes #31 (closed)

Larger-scale testing with the H->ZA code showed some things from !32 (merged) that were not optimal (many things being redefined for the various variations), therefore a few fixes:

• 'systematic' operation nodes behave correctly under caching/redefinition
• function arguments are sorted (alphabetically, by their declaration in the signature) and given standard names - this more or less eliminates function redefinitions (this was #31 (closed))
• a few other oversights/minor fixes

Automatic systematics still bring quite some additional time spent in the python part (Selections affected by many systematic variations are fairly expensive; adding plots to existing selections scales fairly well), more structural solutions are tracked in #34 (closed) .