Decouple root dependencies by using uproot and scipy stats

[kratsg]

• https://github.com/cranmer/fast-asimov-utils
• https://gitlab.cern.ch/lgagnon/mbj_nn/blob/master/python/mbj_nn/utils.py#L165

def significance(signalExp, backgroundExp, relativeBkgUncert):
    """ Numpy/Scipy port of the RooStats function `BinomialExpZ'

    See: https://root.cern.ch/doc/master/NumberCountingUtils_8cxx_source.html
    """
    # pylint: disable=invalid-name
    mainInf = signalExp + backgroundExp
    tau = 1.0 / backgroundExp / (relativeBkgUncert * relativeBkgUncert)
    auxiliaryInf = backgroundExp * tau
    P_Bi = scipy.special.betainc(mainInf, auxiliaryInf + 1, 1.0 / (1.0 + tau))
    return - scipy.special.ndtri(P_Bi)

[kratsg]

I've decoupled RooStats with 3acfd8b.

[kratsg]

I've decoupled most of ROOT (except for plotting) from 3acfd8b to e1c37a8.

e1c37a...3acfd8

[kratsg]

Closing this for now as most of it is decoupled except for plotting which isn't under the scope of this issue.