__array_wrap__ broken for insert, etc.

[dopplershift]

The following code:

import numpy as np
import pint
ureg = pint.UnitRegistry()
a = np.arange(3) * ureg.meter
np.insert(a, 2, 4 * ureg.meter)

produces:

~/miniconda3/envs/py36/lib/python3.6/site-packages/numpy/lib/function_base.py in insert(arr, obj, values, axis)
   4439         new[tuple(slobj)] = arr[tuple(slobj2)]
   4440         if wrap:
-> 4441             return wrap(new)
   4442         return new
   4443     elif indices.size == 0 and not isinstance(obj, np.ndarray):

~/miniconda3/envs/py36/lib/python3.6/site-packages/pint/quantity.py in __array_wrap__(self, obj, context)
   1342 
   1343     def __array_wrap__(self, obj, context=None):
-> 1344         uf, objs, i_out = context
   1345 
   1346         # if this ufunc is not handled by Pint, pass it to the magnitude.

TypeError: 'NoneType' object is not iterable

I have no idea if insert has any hope of working right now, but that exception comes from code that is obviously wrong:

pint/pint/quantity.py

Lines 1428 to 1430 in 241bf7d

	def __array_wrap__(self, obj, context=None):
	uf, objs, i_out = context

Here context defaults to None (as is documented for numpy), but the first line in __array_wrap__ tries to unpack it unconditionally. I have no idea what the right behavior is in this case.

[hgrecco]

Revisit after #905

[jthielen]

Based on https://docs.scipy.org/doc/numpy-1.17.0/reference/arrays.classes.html#numpy.class.__array_ufunc__, it looks like __array_wrap__ is disabled since __array_ufunc__ is defined (thus why __array_wrap__ has been removed in #905), so I don't think this problematic code is reached even now?

np.insert would be a good __array_function__ implementation to add either in #905 or a follow-up.