Probes do not store all oversampling information

[hoogerheide]

This issue is actually two-fold:

1. General Probe objects (e.g. NeutronProbe) store oversampling information, but only general oversampling, not oversampling created with the critical_edge function.
2. QProbe (frequently used with data from facilities that do not produce dTheta / dLambda data explicitly from reduction) does not store oversampling information at all.

The impact of this is that oversampling information is lost when deserializing a serialized model (e.g. from a session history), and the fit results are different than the saved value.