Improve toys
Bug fix, reported by @ivnoviko
Make sure that const, poisson, and multiplicative uncertainties are scaled corresponding to modified values of the central value. I.e. for const (multiplicative) errors absolute (relative) uncertainties are preserved.
Introduce additional warning for negative pseudo-data, reset poisson errors to constant if occurs.
Can be tested using attached steering. The signature of the fix that uncertainties reported in fittedresults.txt, for jobs without minimization (same theory), should remain invariant vs ISeedMC variation, for each combination of valid Chi2Settings settings.
[steering.txt] (/uploads/d4f47b39f2797e3289c4c92d98fc859a/steering.txt)