Thanks to this registration to process_group_pid_set allows cleanup of test
suite too.

If something else is left behind, those are disconnected processes of test
suite itself, which shall be cleared in that one.

For now move supervisord_pid_file to global and calculate it "magically".
Thanks to this it is proven, that supervisord can be killed correctly.

In case if something went wrong in previous run, one can have problem with
partition and its permission can be wrong.

Avoid using not required extensions:

 * slapos.rebootstrap is not needed, as python running buildout is somehow
   trusted, and it is enough to use python section in order to have own
   python compilation used by running software
 * slapos.zcbworkarounds is useful only in corner cases, which are not
   introduced in this profile
 * mr.developer is not used at all

As soon as slapos.core (slapgrid) has zc.buildout in its search path it is
using this one. And thanks to pinning to SlapOS patched buildout and adding
slapos.libnetworkcache networkcache will become enabled on all software runs.

It allows to use latest code version without publishing to pypi.

lxml is available on pypi so direct there.

Avoid using any meta-information, as they direct to lxml.de site, which is
down really often.

Implement a early draft way to deploy a cluster with 5 zopes:

 1 for distribution
 2 for activities
 2 for User usage with HAProxy.

http://sourceforge.net/projects/freetype/files/freetype2/2.4.5/README/view