memcached should detect and abort stuck commands

As seen in MB-56970 a command could enter an endless loop and cause additional problems (like MB-56972).

Memcached should detect if a command enters an endless loop and terminate the command if it is stuck.

In its simplest form we could set a max number of times the command may return ewouldblock. Note that we have some commands which may return ewb a lot of times. rangescans for instance so we would need some sort of "api" to make sure that the command has any progress.