[Bug][Numpy] Test fails for sinc overloading

[fjosw]

Our test for applying sinc to an Obs is failing with numpy 2.3. I tracked this down to this line which was changed to deal with underflows but does not seem to work anymore with custom object types. We could either address this in numpy directly or find a local fix.

[s-kuberski]

Well, that's annoying. So the reason is that the new piece of code is checking x.dtype and x.dtype.kind which are both not defined for an Obs and would have to be introduced? What would be your opinion on the two choices? I would think that the fix in numpy could also affect other libraries that make use of it such that, in principle, it would be a good thing to fix it...