Memory leak and hang in histogram_bin_edges with strategy doane

For the special case shown below, histogram_bin_edges with strategy doane hangs forever and wastes a single CPU and all system memory.

Reproducing code example:

Attention, will start to eat up all system memory!

import numpy as np
np.histogram_bin_edges([0.0, 0.0, 1.0], "doane", range=(0.0, 1056964608.0))

The error will not appear if only a single 0.0 is passed in the array.

Same thing happens with:

import numpy as np
np.histogram_bin_edges([0.0, 1.0, -1.0], "doane", range=(-1.0, 4294967296.0))

Error message:

No error message, code just hangs and wastes all memory.

Numpy/Python version information:

1.18.1 3.8.1 (default, Jan  8 2020, 23:09:20) 
[GCC 9.2.0]

the same thing happens with:

1.17.4 3.7.4 (default, Nov 21 2019, 11:30:33) 
[GCC 9.2.0]

(Issue found in a unit test using hypothesis)

[charris]

Perhaps it would be possible to warn if it looks like things are not quite right? Not sure what the criterion for that would be.

[languitar]

Adding a max_bins option sounds like something that could also work for the other heuristics. For me it would be ok if the method simply raises an exception for cases where the number of generated bins would exceed max_bins (before actually wasting the memory).