Issue with units in np.log() #2977

Hi,

I've noticed some odd behaviour when running np.log() on Quantity objects. Normally, when I run this on a dimensioned value, I get an error, but I've noticed that if the Quantity._value is of a NumPy type, such as numpy.float64 , then the units seem to be ignored and the log function treats the Quantity as just a normal dimensionless value.

Obviously this can be avoided by just being careful with the units that you're using, but I just thought I'd raise the issue - though this may be more of a NumPy issue, I'm not sure.

Apologies if this has already been raised/resolved elsewhere.

Thanks in advance.

Example (using OpenMM 7.4.2 and NumPy 1.16.4):

>>> x1 = 5.0
>>> print(x1)
>>> np.log(x1)
1.6094379124341003
>>> x2 = 5.0 * angstroms
>>> print(x2)
5.0 A
>>> type(x2._value)
<class 'float'>
>>> np.log(x2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/mls2g13/miniconda2/envs/py36/lib/python3.6/site-packages/simtk/unit/quantity.py", line 202, in __getattr__
    ret_val = getattr(self._value, attribute)
AttributeError: 'float' object has no attribute 'log'
>>> x3 = np.sqrt(25.0) * angstroms
>>> print(x3)
5.0 A
>>> type(x3._value)
<class 'numpy.float64'>
>>> np.log(x3)
1.6094379124341003