ENH: Clarify error message for invalid array indices

[YinonHorev]

This enhancement updates the error message thrown during array indexing
with invalid objects. Previously, the message did not explicitly mention
that tuples of valid indexing objects are also acceptable. This could lead
to confusion, especially when debugging legacy code or for users less
familiar with NumPy's indexing rules.

The new error message now includes "tuples of these objects" to clearly
indicate that tuples containing any combination of valid indices (integers,
slices, ellipsis, numpy.newaxis, and integer or boolean arrays) are valid
indexing methods. This change aims to reduce misunderstanding and improve
the developer experience by providing more direct guidance on valid
indexing types.

See ticket #26115

[mattip]

I am not sure this is an improvement, the message is getting too long. Maybe instead of listing everything that is accepted, reflect that this particular construct is not accepted, and do the C-equivalent of

f"Index error: cannot index with {type(arr)}"

[seberg]

I agree with Matti that it would be cool to print out the type of the object the user was using in the last branch (and possibly such a new branch)!

We could maybe be a bit more specific even? If PyArray_NDIM(arr) != 0 then we assume that something (e.g. a list) got converted to an advanced indexing array. I.e. we can add a second branch here to give an entirely more specific error message, where we know the note added here is more likely to apply.
(It is more like the first branch here that warns about only integer and boolean arrays are supported.)