What to do with NoGIL and Python 3.13 about sherpa

[hamogu]

In the even more unlikely event that we're still here when 3.13. is the minimum version we support....

I think this is one of those cases where we might just want to force the non-threaded version (which is still the default in 3.13 and might be in 3.14?) and deal with it once 3.15 is out and a significant fraction of our users actually use unlicked versions ...

However, that might be good incentive move some of our C code that is in C for historical reasons and not for performance to Python/Numpy (e.g. many of our statistics functions could be one-liners in numpy instead of C implementations - I doubt it changes the runtime significantly, but it's certainly not in the critical path anyway) because then we don't have to do this work - Quansight and other numpy developers will do it for us.

[DougBurke]

In my work in

• #2022

to try and address

• #2007

I have noted that it looks like the non-spawn methods are significantly slower. It'll be interesting (in this unlikely future) to know how that is related to the GIL/NoGIL stuff...