Bug: Histogram2d result does not match that of the numpy function

Thank you for this package, it is indeed very fast and I am looking forward to using it.

There is a discrepancy between your histogram2d function and the numpy equivalent, which is demonstrated below:

```python
import numpy as np
from fast_histogram import histogram2d as fast_hist_2d

n_bins = 5

x = np.array([3,2,5,4,6,2,3,5,3,2,5,3,6,7,3])
y = np.array([1,6,3,2,5,3,1,6,4,6,4,2,4,5,3])

xmin, xmax = np.min(x), np.max(x)
ymin, ymax = np.min(y), np.max(y)

h_fast = fast_hist_2d(x,y,
range=[[xmin,xmax], [ymin,ymax]],
bins = [n_bins,n_bins])

h_np = np.histogram2d(x,y,
range=[[xmin,xmax], [ymin,ymax]],
bins = [n_bins,n_bins])[0]

print "h_fast =\n", h_fast
print "\nh_np =\n", h_np
```
On my machine this outputs

h_fast =
[[ 0. 0. 1. 0. 0.]
[ 2. 1. 1. 1. 0.]
[ 0. 1. 0. 0. 0.]
[ 0. 0. 1. 1. 0.]
[ 0. 0. 0. 1. 1.]]

h_np =
[[ 0. 0. 1. 0. 2.]
[ 2. 1. 1. 1. 0.]
[ 0. 1. 0. 0. 0.]
[ 0. 0. 1. 1. 1.]
[ 0. 0. 0. 1. 2.]]

The fast_histogram result only includes 11 of the 15 points. The 4 missing points are all from the final column

Start by running the reported Python reproduction and inspecting the fast_histogram.histogram2d entry point alongside NumPy's histogram2d behavior. Focus on how values at the maximum range edge are assigned, then add a regression test showing that the fast result matches NumPy and counts all 15 points.

Written by the indexing model from the issue text.