some additional (small) problems with pg_combinebackup and tablespaces

Tomas Vondra pointed out to me a couple of mistakes that I made with
regard to pg_combinebackup and tablespaces.

One is that I screwed up the long_options array by specifying
tablespace-mapping as no_argument rather than required_argument. That
doesn't break the tests I just committed because, in the actual string
passed to getopt_long(), I wrote "T:", which means the short form of
the option works; only the long form does not.

The other is that the documentation says that --tablespace-mapping is
applied to the pathnames from the first backup specified on the
command line. It should have said "final" rather than "first".

Here is a very small patch correcting these regrettable errors.

-- 
Robert Haas
EDB: http://www.enterprisedb.com