Limit maximum number of steps in Subset Simulation

[teobros]

Because of numerical issues with floating point precision, it does not make sense to have estimated probabilities of failure smaller than eps(1). Since the probability of failure of each level is a parameter set by the user, the maximum number of steps to reach this limit pf will change. I sugget to insert a maximum number of step check in subset, blocking the analysis when this number is reached. Such number should be (if I am correct) floor(log10(eps(1))/log10(p0)) As an example, for a p0=0.1 it is 15 steps, and for p0=0.2 is 22 steps

[FriesischScott]

Sounds good, a PR is more than welcome!