lcb_subdoc times out when specifying a durability (other than none).

I noticed this on the transactions codebase, where we do mutate_in operations with a default durability Majority. On a cluster where the bucket that the document in has replicas, or not, it always returns LCB_ERR_TIMEOUT. Though I didn't verify with a C app, I verified with python (though it has a bug where it isn't passing the durability stuff in, I just hacked that in).