[Keep][keep-web] improve observability of buffer cache
The keep-web buffer cache can cause a major performance bottleneck when it is not configured appropriately for the traffic load. This problem is more or less invisible from prometheus because we lack metrics
Add metrics for:
- number of blocks in use in the WebDAVCache
- block churn - can we expose # of blocks that were evicted from the cache while in use? If that's not possible, maybe just the # of cache evictions.
- CollectionBytes in use in the WebDAVCache