convert y to numpy array in _reduce_dimensionality

[stevetracvc]

When using cuml umap, y needs to be a numpy array instead of a list.

Here's another one I didn't check first before fixing 😆

this will fix #1241

I leave y as a list everywhere, since I saw at least a few type hints that specified y could be a list or a numpy array. I don't know the intention there, so I only convert it to a numpy array in the _reduce_dimensionality function instead.

[MaartenGr]

Thanks for this PR! Could you make sure to cast y as a numpy only if y is not None? I believe that since this differs quite a bit from the fix in the issue you mention, it will also affect many other techniques, like (semi-)supervised, guided, and non-y topic modeling. So something like:

y = np.array(y) if y is not None else None
self.umap_model.fit(embeddings, y=y)

Or, we could just use the example as used in the issue you refer. That way, processing of y is contained to a single function only where it is relevant.

[stevetracvc]

So I fixed it in my fork before I saw that discussion, and my thoughts were, _reduce_dimensionality accepts
y: Union[List[int], np.ndarray] = None
so if it's acceptable to pass a list here, then forcing y to an array should occur here. But I just started using this package so I don't know how everything is connected. If you think it's better to move it, let me know.

It passed all tests before I submitted, so I'll see if I can add a test for None here. I didn't actually look to see what tests already exist.

[MaartenGr]

Apologies for the late reply. LGTM, thanks!