Should the fit_method argument be maintained in probability_epidemic()?

[Bisaloo]

Follow up from #77 (comment).

At the moment probability_epidemic() proposes two methods to compute the probability: via optim() and via a grid search.

We should investigate and clarify what the cases where one is better the other is, and document it.

If one option is always better than the other (more performant & more accurate), then it should be presented as the only option to:

• avoid encouraging users to use invalid or deprecated methods
• simplify the function usage
• decrease the maintenance load