This change refactors how watcher manages monkey_patching
modules to achieve 2 goals.
First, we want to ensure the watcher code is tested as it is used
in production. While many tests can run without eventlet,
the existing unit tests depend on eventlet monkey patching
indirectly by importing watcher code that uses eventlet.spawn and
greenthread executors. While that mostly functions today it has
incorrect and inconsistent behaviour on Python 3.9 vs Python 3.12.
Second, the unit tests that test the cmd module were indirectly
monkey patching the test executor during the execution of the tests
as a side effect of importing watcher.cmd. As such the order the tests
execute in and how they are distributed across test workers changed
if the test was monkey-patched or not.
This change makes all tests run with monkey_patching by adding
monkey patching in the watcher/tests/__init__.py
This change also splits the monkey patching from the import
in preparation for an eventual removal of eventlet in a future
release.
Change-Id: I967f3469bd66e69c00863d553bc859343afbb3ff