Clean up the *fit stuff

[vatai]

Ideas:

• write a general Lorentzian_fit
• add others (like Gaussian_fit, quadratic_fit etc)
• unify the use of these *_fit functions

[vatai]

@lmantill says these "fittings" belong to experiments, so e.g. the MARY() experiment could have a fit: bool flag, e.g. `MARY(..., fit=True).

Keep option of doing it manually.

[vatai]

Todo @vatai : compare Bhalf fit (MARY) with multiexp fit (autocorrelation/MC)