Restore numpy compatibility

[MartinBubel]

With the release of numpy>=2, the my_array.data access got removed.

This is used deep inside paramz which makes it incompatible with modern numpy versions.
Since numpy is a core dependency of pretty much anything in scientific computing, this needs to be fixed.

The way the .data accessor was used in paramz is to update the memory location of existing arrays, without re-creating them and without creating temporary data copies.

As off now, it seems like this is something that is not possible anymore in numpy>=2, at least, I could not figure out how - at least for now.

What does this PR do?

• Fixes the package setup
• Fixes testing
• Fix deprecations in numpy and python itself

It may make sense to split this up into multiple PRs later. Sorry for the mess atm.

I have added a _set_mem_addr function in paramz.util.py that serves as a baseline for the problematic update that is still not fixed.
The current variants works with numpy<2 but fails otherwise.

If anyone has got some idea on how to better debug or even fix this, feel free to join :)

Also, this is not based on develop but master. The reason for this is that the latest merge into development seems to have broken something else. We'll fix that later.

[MartinBubel]

Hi @mzwiessele ,
I'm trying to update your lib to numpy>=2 compatibitliy and I'm a little overwhelmed by the complexity of that task atm.

It seems like replacing the .data accessor is becoming really difficult in with numpy>2 these days.
Seems like numpy does not like updating memory addresses of array's data anymore.

Do you have any idea on how we can fix this?