Improve robustness to arbitrary input

I started trying to add some full hypothesis tests such as:

```

# First some tests with hypothesis value generation to make sure our
# implementation doesn't crash, and do basic checks of validity of the output.

@given(values=arrays(dtype='f4', 'f8', '= xmin))
```

but this crashes with a segmentation fault currently, so need to investigate further.

Start with the shown test_1d_basic case and the histogram1d entry point, then run the Hypothesis test to reproduce the segmentation fault. Trace the failing arbitrary input into the implementation and its C extension; done means generated inputs no longer crash and the stated sum assertion passes.

Written by the indexing model from the issue text.