Waste too much time on blob cache operation

[Connor1996]

When doing a long scan with 1000 kvs on master, find it wastes too much CPU time on blob cache mutex and affect scan performance a lot.

But for 4.0, the scan performance is not affected a lot(sorry for the flame graph is missing)

[yiwu-arbug]

By default each cache shard is at least 512KB, which means if cache size is set to 0, there's only 1 shard. It will of course cause mutex contention. Can you try to set cache to a more practical size and see?

https://github.com/facebook/rocksdb/blob/master/cache/sharded_cache.cc#L149

[Connor1996]

Yes, but seems the mutex contention causes too much overhead. And in 4.0 branch, it doesn't have that much overhead when size == 0. I'll take a deep look in the future.