Issue with numba and umap

Hi there

I have an error I'm having trouble resolving. I can't seem to get
numba==0.45.0 and umap-learn<=0.3.8 to work together.

I keep getting the following error:

numba.errors.LoweringError: Failed in nopython mode pipeline (step: nopython mode backend)
Type of #4 arg mismatch: i1 != i32

File "../../miniconda3/envs/solo-sc/lib/python3.7/site-packages/umap/utils.py", line 131:
def make_heap(n_points, size):
    <source elided>
    """
    result = np.zeros((3, int(n_points), int(size)), dtype=np.float64)
    ^

Which seems to be similar to this issue:
https://github.com/lmcinnes/umap/issues/225

I've tried python 3.6.1, 3.6.7, and 3.7 none of which have helped the issue.
Any ideas about troubleshooting would be great. Other versions of Numba and umap-learn woork together for me but for a variety of other dependencies I'd really like the versions mentioned above to work together.

Please let me know if you have any thoughts

Start with the traceback location in umap/utils.py at line 131 and the linked issue #225. Reproduce the failure using Python 3.6 or 3.7 with numba==0.45.0 and umap-learn<=0.3.8, then compare the behavior with the other versions mentioned. Done means identifying whether this version combination is supported and documenting a confirmed resolution or incompatibility.