Improve internal cache of requests

[jolpica]

Currently page cache is done via the cache_page decorator.

However this caches the entire page, and results in separate caches for different formats (e.g. html and json).
Instead the cache should be made on the output of the serializer's to_representation method.

This has the following benefits:

• More reuse of the cache
• Currently when the cache is hit, it does not count towards rate limiting. Which we may not be what we want.