Handling Origin/Cache redirects based on requests from clients and caches

[jhiemstrawisc]

This is linked to issue #436.

This is a nuanced issue; when the client asks the director for a redirect but there are no caches available, we might forward to an origin instead. But it's also valid for an origin to say "I only want to serve my objects to caches" in the event that it has limited resources or has to worry about egress costs. Is this something the Director should be aware of? Maybe origins should be responsible for vetting that the client requesting an object is actually a cache registered in the federation? In my view of things, every type of resource owner should be able to specify exactly how they want their resources to be used. That would mean:

1. Origin owners should be able to specify "Don't cache my stuff in the federation," "Only access my stuff through a cache in the federation," or "Only caches X,Y,Z are allowed to access my stuff"
2. Cache owners should be able to specify things like "I only want to cache objects from namespaces X,Y,Z" or "I want to check 3 other caches before I pull from an origin."