Add metadata plugin mechanism

Until now we only supported storing resources in memory, and that
mechanism was tightly coupled with everything else.

This patch decouples the metadata storing into a plugin mechanism using
'Python entrypoints.  The default storing mechanism is still in memory,
and for now it's the only mechanism available.

The JSON serialization mechanism has not yet been adapted to the
metadata plugin mechanism, so it will only work with the default plugin.

setup.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python
 # -*- coding: utf-8 -*-
 
-from setuptools import setup
+import setuptools
 
 with open('README.rst') as readme_file:
     readme = readme_file.read()
@@ -47,7 +47,7 @@ extras = {
                   'infi.dtypes.iqn'],
 }
 
-setup(
+setuptools.setup(
     name='cinderlib',
     version='0.1.0',
     description=("Cinder Library allows using storage drivers outside of "
@@ -56,11 +56,7 @@ setup(
     author="Gorka Eguileor",
     author_email='geguileo@redhat.com',
     url='https://github.com/akrog/cinderlib',
-    packages=[
-        'cinderlib',
-    ],
-    package_dir={'cinderlib':
-                 'cinderlib'},
+    packages=setuptools.find_packages(exclude=['tmp', 'tests.*', 'tests.*']),
     include_package_data=True,
     install_requires=requirements,
     extras_requires=extras,
@@ -81,4 +77,9 @@ setup(
     ],
     test_suite='unittest2.collector',
     tests_require=test_requirements,
+    entry_points={
+        'cinderlib.persistence.storage': [
+            'memory = cinderlib.persistence.memory:MemoryPersistence',
+        ],
+    },
 )