🐛 optimize.least_squares: Allow scalar floating point return type in residual function

[haarisr]

According to the docs, the least squares residual function can return a 1D array or scalar.

Added a test as well for the change

[github-actions]

mypy_primer results

No ecosystem changes detected ✅

[jorenham]

Thanks for the PR!

Could you provide a (runtime) example of such usage? Just so that I understand a bit better what's going on.

edit: after looking a bit closer I realized this fixes a pretty obvious issue; so no need for that example

[jorenham]

Also, do you have any idea if this also is the case for the other callables (jac and loss)?

[jorenham]

I've got nothing to complain here, and I really like the testcase; it's nice and robust.

So unless the jac or loss also need to be fixed, this should be good to merge, AFAIK.

[haarisr]

So unless the jac or loss also need to be fixed, this should be good to merge, AFAIK.

It doesn't. jac needs to be at_least2d and if loss is a callable, it need to be of shape 3xM

[jorenham]

Ok, in it goes then. Thanks @haarisr !