ovsdb: Allow recovery from transient write errors in log implementation.

ovsdb: Allow recovery from transient write errors in log implementation.

Until now, the OVSDB data journaling implementation has made write errors
"sticky", so that a single write error persists as long as ovsdb-server is
alive.  However, some kinds of write errors (such as ENOSPC) can be
transient in practice.  I don't know of a good reason to make such errors
sticky, so this commit makes the journaling code retry writes even after
an error occurs, allowing ovsdb-server to recover from transient errors.

Reported-by: likunyun <kunyunli@hotmail.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Ethan Jackson <ethan@nicira.com>