NumPy exp and log keeps the type

Previous behaviour: when it`s passed a dimensionless quantity, it does
not return a dimensionless quantity; it returns a standard numpy array
(or scalar)

New behaviour: returns a dimensionless quantity.

Breaking Change: to regain the old behaviour, get the magnitude of the
resulting value. e.g.

>>> np.log(5 * ureg.dimensionless).magnitude

Close #95

pint/quantity.py

@@ -1052,6 +1052,7 @@ class _Quantity(SharedRegistryObject):
     #: will set on output.
     __set_units = {'cos': '', 'sin': '', 'tan': '',
                    'cosh': '', 'sinh': '', 'tanh': '',
+                   'log': '', 'exp': '',
                    'arccos': __radian, 'arcsin': __radian,
                    'arctan': __radian, 'arctan2': __radian,
                    'arccosh': __radian, 'arcsinh': __radian,

pint/testsuite/test_issues.py

@@ -530,3 +530,8 @@ class TestIssuesNP(QuantityTestCase):
                          '1 count')
         self.assertEqual('{0:~}'.format(1 * self.ureg('MiB')),
                          '1 MiB')
+
+    def test_issue482(self):
+        q = self.ureg.Quantity(1, self.ureg.dimensionless)
+        qe = np.exp(q)
+        self.assertIsInstance(qe, self.ureg.Quantity)