uint64 converted silently to float64 when adding an int

This code

```
import numpy
a = numpy.zeros(1, dtype = numpy.uint64)[0]
print(type(a))
i = 1
print(type(i))
a += i
print(type(a))
```

prints

```
<class 'numpy.uint64'>
<class 'int'>
<class 'numpy.float64'>
```

which was a big surprise for me. Why would adding an integer to uint64 result in a floating point value?
