Spring Use Collection View instead of a separate collection

• return a keySet view of map instead of cacheNames
• remove cacheNames
• simplify get-if-put -> computeIfAbsent
• remove synchronized block

Comment From: jhoeller

The main reason for the separate cacheNames collection is the ordering of cache names which can be important for startup-time cache registration at least, not so much for runtime-added caches. We generally recommend startup-time cache definitions, so it seems sensible to preserve the current arrangement.

If we were to give up on cache name ordering, your PR would make sense, of course. Thanks for your efforts, in any case!

Comment From: bananayong

Thank you for your kind explanation! I missed javadoc of updateCacheNames method.