notnull very slow compared with 0.3.0

[wesm]

from Rory Grandin on the pystatsmodels mailing list

I'm not sure if anyone else has come across this, but I use notnull
extensively for checking if individual values are valid, and when
switching from pandas 0.3 to 0.4, noticed an almost fivefold increase
in the runtime for notnull.  I think this is because notnull now calls
isnull, thus adding an additional burden.  I've managed to reduce the
overhead somewhat by using the recommended lazy importing structure
from the performance tips on the python wiki (http://wiki.python.org/
moin/PythonSpeed/PerformanceTips#Import_Statement_Overhead).  The call
to isnull now takes about a third of the time it was before.

Please let me know if you need any more information, I'm a bit new to
this, and would like to help out in any way I can :)