Rationalize and 'fix' memcached / ep_engine logging

The log levels used for things in memcached and ep_engine are extremely inconsistent, and should be fixed.

For example, we have four log levels: Warning, Info, Debug and Detail. The default level appears to be Warning, therefore people have (IMHO incorrectly) made everything they want to see in the log by default as Warning, when they aren't actually things to warn about. A few examples:

These should be INFO:

Wed Aug 20 17:42:15.515493 BST 3: (No Engine) Bucket default registered with low priority
Wed Aug 20 17:42:15.515525 BST 3: (No Engine) Spawning 4 readers, 4 writers, 1 auxIO, 1 nonIO threads
Wed Aug 20 17:42:15.999738 BST 3: (default) metadata loaded in 483 ms
Wed Aug 20 17:42:15.999766 BST 3: (default) Enough number of items loaded to enable traffic
Wed Aug 20 17:42:15.999927 BST 3: (default) warmup completed in 483 ms

These should be INFO at most, maybe even DEBUG:

Wed Aug 20 17:52:23.692284 BST 3: (default) DCP (Consumer) eq_dcpq:replication:n_0@192.168.0.86->n_1@127.0.0.1:default - (vb 1023) Attempting to add stream with start seqno 0, end seqno 18446744073709551615, vbucket uuid 250149181088775, snap start seqno 0, and snap end seqno 0
Wed Aug 20 17:52:23.709485 BST 3: (default) DCP (Consumer) eq_dcpq:replication:n_0@192.168.0.86->n_1@127.0.0.1:default - (vb 1023) Attempting to add takeover stream with start seqno 0, end seqno 18446744073709551615, vbucket uuid 140850177473972, snap start seqno 0, and snap end seqno 0

(there are plenty more examples)...

If we want our default level to be lower (and log more detail by default) then that's probably a fair thing to do, but we shouldn't just dump everything under WARNING as that basically defeats the point of different log levels.