Update numpy req to 2

[paulf81]

Update numpy req to 2

Testing an update based on #1030 where the numpy requirement is pushed up to compatiable version of 2.0.

An alternative approach would be to span version 1 and 2 using > and < operators, but I understand this to be non-preferred in semantic versioning, since testing would only apply to newer major version, and so the previous major version is not truly supported.

Using this PR also to check if having the requirement at ~=2.0 raises any incompatibilities.

Related issue

#1030

[misi9170]

I'm OK with going up to ~=2.0. We'll also be dropping support for python 3.8 in the next minor FLORIS version #1042 , and this feels like a similar thing. However, there are likely to be users that will still be on numpy 1.x, who may be prevented from upgrading to FLORIS 4.3 if we introduce this requirement.

[thebaptiste]

May be you could check it works with numpy 2 (and fix incompatibilities if necessary) but set numpy requirement to numpy>=1.20 or something like numpy>=1.20,<2.1 (if its possible to have a code working with both numpy 1.20 and numpy 2.0... or even better 2.1 or 2.2 !) in order to let users choose between numpy 1.20 and numpy 2.0 ?

Dependencies with ~= will often be a problem for users if they want to use together floris and other packages... and in any case it's an investment in the future for a user or a project to move from numpy 1 to numpy 2 !

[thebaptiste]

Here is another problem with requirement numpy~=1.20 : if you want to use together both floris and numpy 2.0 (even numpy 2.2), it's possible ! you have to use... floris 3.5 ! because in floris 3.5 the requirement was numpy>=1.20. Awful isn't it ? (it probably doesn't work but it's possible to install !). Try pip install floris==3.5.

I think there will be more difficulties for the majority of users if you do not upgrade to numpy 2.x than if you upgrade to numpy 2.x
There are less and less packages not compatible with numpy 2.0, Floris is one of them...
I think it's better to encourage users to upgrade to numpy 2 (for their own good) if they want to use new floris functionalities than to keep them in their "dead end" with numpy 1.20...

[rafmudaf]

Some things to consider:

• I see 2.1 drops support for Python v3.9 (https://numpy.org/devdocs/release/2.1.0-notes.html), how does that impact the Numpy version that FLORIS should support?
• What's the impact to performance? Reading through the Numpy release notes, I think moving to Numpy v2 will significantly improve performance, but it would be good to quantify this
• It's worth double checking that adjacent tools in the WETO Stack also support Numpy v2 (FLASC, OpenOA, anything else that integrates FLORIS -> FLOWERS?)

[paulf81]

Ok, I think we're in agreement that we should make the move up to numpy 2, but after some discussions we'd like to do things in particular order. I will open a separate discussion to track the sequence (#1056)