Last time I had this problem to solve, the difficulty wasn't finding an SA, it was finding a reasonably priced, working, calibrated LISN. I suggest making sure you have a good solution for this first.
You don't need high bandwidth for PSU noise measurements, and for qualitative measurements, any working SA at all is probably OK. If your design is a bit close to the limit, then you might want to find one with a quasi-peak detector, and you might want to look at getting your complete set-up calibrated.
Something like a used HP 8594E could be a good buy.