Also begin to work on rpms

OSDs shouldn't crash or hang with long iodepths anymore

Reply -EPIPE for non-existing pools because we assume that it means
that pool config isn't loaded yet. Previously OSD crashed on such operations

There was a FIXME and I actually hit it during tests :)

This is the new guaranteed unblocking method which replaces old trims
in init and rollback, and also fixes a possible stall when just several
writes in the beginning of the journal are flushed without triggering
a subsequent trim.

1) Update journal's used_start in memory only after updating journal superblock.
Doing the opposite is incorrect because part of the journal will be lost if writers
overwrite its old beginning.

2) Sync journal device after updating the superblock.

3) Do not trim in rollback and init because trimming there would also require
updating the superblock. And the only reason to trim in both those places was
to unblock writers. And a guaranteed unblocking method will follow in the next
commit :)

Establish watcher connection after loading PGs

Also fix possible race condition which could in theory lead to "command out of sync"
and a buffer overflow that could happen on incorrect server response.