casting to quandtype emits uneccesary warnings

[mattip]

When using np.astype and the array has inf or nan, I see a RuntimeWarning: invalid value encountered in cast. On internal numpy float dtypes these are not emitted.

Run with python -Walways

import numpy as np
import numpy_quaddtype
a = np.arange(30, dtype=np.float32); 
a[-1] = np.inf
q = numpy_quaddtype.QuadPrecision
d = np.float64
warns = a.astype(q)
# <stdin>:1: RuntimeWarning: invalid value encountered in cast
nowarns = a.astype(d)
# This does not warn
warns.astype(d)

I think this is coming from sleef?

[SwayamInSync]

Yes it seems to be coming from sleef and also comes in during extreme cases comparison using sleef specific operations. Not 100% sure but its possible sleef using quiet comparison routines.
@seberg suggested this flag -ftrapping-math in this comment #121 (comment) (It is not yet applied in the sleef's meson build) I'll look into it