default rng_med_win in lalpulsar_ComputeFstatMCUpperLimit

@karl-wette @keith-riles I noticed that this code has a default of 50 for the running median window (see here), while all other F-stat codes use 101 (as formalised via the FstatOptionalArgsDefaults struct).

It should only have a small impact, and the code's outputs are only generally used as starting guesses for more involved UL procedures as far as I'm aware (I've included Keith due to the SNR Weave usage). But it would probably be a good idea to change this for consistency. Any objections based on reproducibility of previous results? Though as it usually is just a starting guess, I hope this won't be much of an issue.