Handle undefined correction factors

[eseiler]

For certain combinations of parameters, the fp_correction will yield inf.
E.g. for h=2, fpr=0.1, the function collapses at around 355 TB: WolframAlpha Plot

Idea:

• Set f_h to a high number to discourage splitting (inf is actually a high number :))

[feldroop]

Do we need to do anything, if the fp_correction already yields inf for the problematic cases?

[eseiler]

Not really, but it may cause trouble in the future.
It seems to be more a problem of x^y being 0 for any positive x < 1 and large enough y, and not so much that there is no correction factor.

But I haven't looked into it yet.