Enforces that stacklevel is used in warnings.warn and "ignore"
is not used in filterwarnings or simplefilter for most of numpy.

Reloading currently causes problems because global classes defined in
numpy/__init__.py change their identity (a is b) after reload. The
solution taken here is to move those classes to a new non-reloadable
module numpy/_globals and import them into numpy from there.

Classes moved are ModuleDeprecationWarning, VisibleDeprecationWarning,
and _NoValue.

Closes #7844.

There seems to be little use in reloading numpy as any changed modules
that are imported into numpy would need to be reloaded first in order to
see any changes. Furthermore, reloading causes problems as global
classes defined in numpy/__init__.py change their identity. Hence we
raise a RuntimeError when an attempt to reload numpy is made.

Closes #7844.

Fix #2303
Cache queries wont miss because the whole tuple is cached as key
and not just dtype
Adapted from patch submitted by Colin Hogben

When you run 'python -V' under Anaconda, it
returns for example, 'Python 3.4.3 :: Continuum
Analytics, Inc.' However, the original parsing
of the version in 'test_f2py' assumed there was
nothing following the version number, causing a
ValueError because you can't assign three
variables to four components that you get from
splitting on the '.'

The 'sys.executable' can come in various names, but
the three main ones are "python", "python{major_version}",
and "python{major_version.minor_version}". The current
version of the f2py test assumes that only the latter
two are used. Since "f2py" is generally versioned,
using the executable basename "python" will make it
impossible to find. This commit fixes that issue
by using a sure-fire way of getting the Python version.

Fixed test_scripts.test_f2py test so that
it can pass correctly on win32 virtualenvs,
in which the Python executable and the
f2py.py file are both in the Scripts
directory.

Addresses comments of @pv on gh-6895.