Fix broadcasting scale arg of entropy

[toslunar]

Fix #8931

This PR also fixes the following error with numpy<1.12. (numpy.place does not reject >= 2-d vals)

>>> stats.norm(loc=0, scale=np.array([[1, 2], [3, 4]])).entropy()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/tos/py3sandbox/lib/python3.6/site-packages/scipy/stats/_distn_infrastructure.py", line 499, in entropy
    return self.dist.entropy(*self.args, **self.kwds)
  File "/Users/tos/py3sandbox/lib/python3.6/site-packages/scipy/stats/_distn_infrastructure.py", line 1119, in entropy
    place(output, cond0, self.vecentropy(*goodargs) + log(scale))
  File "/Users/tos/py3sandbox/lib/python3.6/site-packages/numpy/lib/function_base.py", line 2002, in place
    return _insert(arr, mask, vals)
ValueError: object too deep for desired array

[ev-br]

Nice!

This also needs a test. Here's one, https://github.com/scipy/scipy/compare/master...ev-br:pr/8932?expand=1
can you take it over?

Once we go down that route, there's one more bug:

In [2]: from scipy import stats

In [3]: stats.norm.entropy(scale=2)
Out[3]: array(2.11208571)

In [4]: stats.norm.entropy(scale=[2])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-4-e060df9994ec> in <module>()
----> 1 stats.norm.entropy(scale=[2])

~/repos/scipy/build/testenv/lib/python3.5/site-packages/scipy/stats/_distn_infrastructure.py in entropy(self, *args, **kwds)
   1111         # NB: for discrete distributions scale=1 by construction in _parse_args
   1112         args = tuple(map(asarray, args))
-> 1113         cond0 = self._argcheck(*args) & (scale > 0) & (loc == loc)
   1114         output = zeros(shape(cond0), 'd')
   1115         place(output, (1-cond0), self.badvalue)

TypeError: unorderable types: list() > int()

This is due to a missing loc, scale = map(asarray, (loc, scale)) in the entropy method.

[toslunar]

Thanks! I merged your commit, made the fix you mentioned, and added a test.

[ev-br]

LGTM, merging. Thanks @toslunar!

It might make sense to do similar things for other methods of distributions, is it something you'd be interested in?