Improving monitoring of cache misses

[bbockelm]

The monitoring loop on the cache currently writes files into the cache then checks to ensure it can read them out via XRootD, effectively testing out the "cache hit" use case.

Fabio reports that there are use cases where the caches fail but only for "cache misses". Hence, we need to extend the monitoring to work on this case as well.

Here's the general concept I had in mind:

1. Add functionality to the director that causes it to export a "fake" part of the namespace, solely for monitoring. No XRootD server involved, just the director redirecting to itself and generating the response on demand. From experience: make the response a few dozen bytes; make the name unique per request and not all in the same "directory"; add a rate limiter to prevent from abuse.
2. The cache periodically downloads a generated object name from its own XRootD endpoint using a signed token. This will always have a cache miss, go to the director, and the director will internally redirect to itself (serving up the requested object).