pg_rewind: Don't error if the two clusters are already on the same timeline

pg_rewind: Don't error if the two clusters are already on the same timeline

This previously resulted in an error and a nonzero exit status, but
after discussion this should rather be a noop with a zero exit status.

This is a back-patch of commit 6b34e5563849edc12896bf5754e8fe7b88012697,
plus two changes from commit e50cda78404d6400b1326a996a4fabb144871151
that teach pg_rewind to allow the initial control file states to be
DB_SHUTDOWNED_IN_RECOVERY as well as DB_SHUTDOWNED.  That's necessary
to get the additional regression test case to pass, and the old behavior
seems like rather a foot-gun anyway.

Peter Eisentraut and Tom Lane