adds ability to pass kwargs to guess_param_defaults in Fitter

[cosama]

This quick fix, adds the ability to call

fitter.guess_param_defaults(...)

before fitting a module to change the default parameters of the guess. Parameters can be passed to the various components simply by referencing them with their prefix. Calling the following code:

import becquerel as bq

model = bq.fitting.GaussModel

fitter = bq.Fitter(
    model,
    ...
)

fitter.guess_param_defaults(gauss_width_ratio=0.05, update=True)

fitter.fit(backend="lmfit-pml")

for example will use the proper width_ratio in the model.guess(...) function, instead of the default value as done right now.

The current behavior -- ignoring all kwarg arguments -- looks like a bug to me.

Additionally I also added the ability to the ExpGauss model to create a left tailing distribution, by selecting gamma as negative. So far we only were able to define right tailing distributions.

[markbandstra]

Please add a test or two to cover these changes.