Fallback to Model Permissions (Core)

[doganmeh]

Related to #380, and #49.

This updates core.py and test_core.py to allow the following checker functions optionally fallback to (or include) model level permissions:

get_user_perms(self, obj, fallback_to_model=None)
get_group_perms(self, obj, fallback_to_model=None)
get_perms(self, obj, fallback_to_model=None)

Please note the fallback_to_model parameter.

This also includes, a global setting two that end: FALLBACK_TO_MODEL. Function level takes precedence if both provided. Overall defaults to False in order not to change the default behavior.

When get_perms is used with fallback_to_model=True, the combined permissions go through the existing caching mechanism, so a performance degradation is not expected (although not tested).

Tests are implemented, but documentation is not. More is planned.

[doganmeh]

previous commit adds standalone cache for model level permissions even if directly accessed via get_user_perms and get_group_perms

[doganmeh]

I now more firmly believe that falling back to model permissions should be via falling back to django default backend, not within Guardian. While this is in the mean time (that was what it was meant to be) useful; should eventually be get rid of. I also have a pull request at Django to help this. Maybe it is better not to pull this in, in hopes that, that one will be.