rm -Rf $BUILD_DIR/outputfs if it is not mounted anymore

Instead of special code in cleanup_build_area_and_end_program(), use the
shared helper to remove $BUILD_DIR/outputfs. This translates to using
rmdir always instead of using rm -Rf in case the filesystem is not
mounted. This is simpler and safer (disaster could arise if the
mointpoint command fails for any unexpected reason, for example).

It intentionally reverts PR rear#578 (commit
606d8ce). It is not clear why was
.lockfile present in the unmounted outputfs (this was the reason for the
change). According to my understanding of the code, this should not
happen. Even if this happens, it means that some part of the code is
using outputfs incorrectly (it should be used only if mounted, otherwise
data put there are lost) andf it is better to be warned about the
problem rather than ignore it by blindly removing the whole directory.

Instead of special code in cleanup_build_area_and_end_program(), use the
shared helper to remove $BUILD_DIR/outputfs. This translates to using
rmdir always instead of using rm -Rf in case the filesystem is not
mounted. This is simpler and safer (disaster could arise if the
mountpoint command fails for any unexpected reason, for example).

It intentionally reverts PR rear#578 (commit
606d8ce). It is not clear why was
.lockfile present in the unmounted outputfs (this was the reason for the
change). According to my understanding of the code, this should not
happen. Even if this happens, it means that some part of the code is
using outputfs incorrectly (it should be used only if mounted, otherwise
data put there are lost) and it is better to be warned about the
problem rather than ignoring it by blindly removing the whole directory.

Instead of special code in cleanup_build_area_and_end_program(), use the
shared helper to remove $BUILD_DIR/outputfs. This translates to using
rmdir always instead of using rm -Rf in case the filesystem is not
mounted. This is simpler and safer (disaster could arise if the
mountpoint command fails for any unexpected reason, for example).

It intentionally reverts PR rear#578 (commit
606d8ce). It is not clear why was
.lockfile present in the unmounted outputfs (this was the reason for the
change). According to my understanding of the code, this should not
happen. Even if this happens, it means that some part of the code is
using outputfs incorrectly (it should be used only if mounted, otherwise
data put there are lost) and it is better to be warned about the
problem rather than ignoring it by blindly removing the whole directory.

Instead of special code in cleanup_build_area_and_end_program(), use the
shared helper to remove $BUILD_DIR/outputfs. This translates to using
rmdir always instead of using rm -Rf in case the filesystem is not
mounted. This is simpler and safer (disaster could arise if the
mountpoint command fails for any unexpected reason, for example).

It intentionally reverts PR rear#578 (commit
606d8ce). It is not clear why was
.lockfile present in the unmounted outputfs (this was the reason for the
change). According to my understanding of the code, this should not
happen. Even if this happens, it means that some part of the code is
using outputfs incorrectly (it should be used only if mounted, otherwise
data put there are lost) and it is better to be warned about the
problem rather than ignoring it by blindly removing the whole directory.

…-updated

Fix backup removal in exit task and cleanup handling of URL mountpoints:
Cleanup of temporary mount point handling, particularly for output.
Unification of mount point umount and cleanup move
to the mount_url() and umount_url() functions.
Replaced the various "rm -rf" of the mountpoint by "rmdir" which fixes
#2611
Added lazy umount in case normal umount does not succeed.
If build dir is kept (cf. KEEP_BUILD_DIR), propose a safe way to remove it
to the user via "rm -Rf --one-file-system" instead of just "rm -Rf" where
the user risks to remove everything below that mountpoint if still mounted.
Fixes also some other bugs noted in the process:
Filesystem-specific umount command not called
20359a9#r51319634
Unknown schemes considered invalid, see the discussion under
#932
Identical scripts under DUPLICITY and YUM replaced by symlinks.
Reverted
#782
that had reintroduced
#465
which got re-reported as
#2611
Reverted
#578
because it is not clear how .lockfile can exist in the
unmounted filesystem, and if it does, it is a bug.
Reverted 
d850c40
because it became meanwhile obsoleted by
a8fdc44

Instead of special code in cleanup_build_area_and_end_program(), use the
shared helper to remove $BUILD_DIR/outputfs. This translates to using
rmdir always instead of using rm -Rf in case the filesystem is not
mounted. This is simpler and safer (disaster could arise if the
mountpoint command fails for any unexpected reason, for example).

It intentionally reverts PR rear#578 (commit
606d8ce). It is not clear why was
.lockfile present in the unmounted outputfs (this was the reason for the
change). According to my understanding of the code, this should not
happen. Even if this happens, it means that some part of the code is
using outputfs incorrectly (it should be used only if mounted, otherwise
data put there are lost) and it is better to be warned about the
problem rather than ignoring it by blindly removing the whole directory.