Refactoring of compression options in pg_basebackup

Hi all,
(Added Georgios in CC.)

When working on the support of LZ4 for pg_receivewal, walmethods.c has
gained an extra parameter for the compression method.  It gets used on
the DirectoryMethod instead of the compression level to decide which
type of compression is used.  One thing that I left out during this
previous work is that the TarMethod also gained knowledge of this
compression method, but we still use the compression level to check if
tars should be compressed or not.

This is wrong on multiple aspects.  First, this is not consistent with
the directory method, making walmethods.c harder to figure out.
Second, this is not extensible if we want to introduce more
compression methods in pg_basebackup, like LZ4.  This reflects on the
options used by pg_receivewal and pg_basebackup, that are not
inconsistent as well.

The attached patch refactors the code of pg_basebackup and the
TarMethod of walmethods.c to use the compression method where it
should, splitting entirely the logic related the compression level.

This is one step toward the introduction of LZ4 in pg_basebackup, but
this refactoring is worth doing on its own, hence a separate thread to
deal with this problem first.  The options of pg_basebackup are
reworked to be consistent with pg_receivewal, as follows:
- --compress ranges now from 1 to 9, instead of 0 to 9.
- --compression-method={none,gzip} is added, the default is none, same
as HEAD.
- --gzip/-z has the same meaning as before, being just a synonym of
--compression-method=gzip with the default compression level of ZLIB
assigned if there is no --compress.

One more thing that I have noticed while hacking this stuff is that we
have no regression tests for gzip with pg_basebackup, so I have added
some that are skipped when not compiling the code with ZLIB.

Opinions?
--
Michael