annotation resource caching should be handled by the resolver

[balmas]

I think the resolver should be responsible for determining whether and how an annotation resource is cached, rather than this being baked into the AnnotationResource.

E.g. we currently have:

def read(self):
        """ Read the contents of the Annotation Resource

        :return: the contents of the resource
        :rtype: str or bytes or flask.response
        """
        if not self.__content__:
            self.__retriever__ = self.__resolver__.resolve(self.uri)
            self.__content__, self.__mimetype__ = self.__retriever__.read(self.uri)
        return self.__content__

This is problematic in a variety of ways ...

1. the memory consumption could get quite huge as all annotation resources will be cached in memory after first read
2. if the annotation resource itself is not static, such as might be the case with a remote resource served via HTTP, there is no way for it to be refreshed

Probably the resolver should be configured with a cache provider, and it should be up to the resolver to figure out when to refresh the cache (e.g. for an http retriever via cache headers, for a local resource by date/timestamp/file size, etc.)

[PonteIneptique]

Agree with the general idea, resolver should have the ability to get a Cache instance fed up to them on instantiation. Great idea.

As for the RAM issue, it is not totally correct. What you describes is the case in a non quite developy environment, where there is one thread which never sleeps. Basically, bad configs (heroku) or flask default .run(). I do not think this holds true for balanced / right envs