ENH: add support for `unravel_index`

[OmarManzoor]

Would it be worth adding support for

• unravel_index - numpy unravel_index

I encountered this in the following PR: https://github.com/scikit-learn/scikit-learn/pull/33906/changes

CC: @lucascolley

[lucascolley]

there is already an issue for swapaxes at gh-97. The key question seems to be whether we want to include the implementation at #97 (comment) in array-api-extra, or extend the spec so that this is a one-liner with moveaxis: #97 (comment)

[lucascolley]

could you research what torch and jax have for unravel_index?

[OmarManzoor]

""" 
Converts a tensor of flat indices into a tuple of coordinate tensors that index into an arbitrary tensor of the specified shape.
"""
torch.unravel_index(indices, shape)

• https://docs.pytorch.org/docs/2.12/generated/torch.unravel_index.html

"""
Convert flat indices into multi-dimensional indices.

JAX implementation of [numpy.unravel_index()](https://numpy.org/doc/stable/reference/generated/numpy.unravel_index.html#numpy.unravel_index). The JAX version differs in its treatment of out-of-bound indices: unlike NumPy, negative indices are supported, and out-of-bound indices are clipped to the nearest valid value.
"""
jax.numpy.unravel_index(indices, shape)

• https://docs.jax.dev/en/latest/_autosummary/jax.numpy.unravel_index.html

Similar signatures to numpy

[lucascolley]

seems like a reasonable addition. Might be worth raising on the array-api repo too.