Because dufs only reads certificate on startup it does not detect when
the certificate was renewed.

The only thing supported by dufs is to restart, but we can not restart
the service, because the partition can not communicate with the
supervisor.
To solve this, use another level of process supervisor: using zdaemon
we can run a process in the foreground, so that it stays under
slapos' supervisor control, but still expose a way to restart the
process using zdaemon API.

This reverts commit b2db18a0cb672204ff1dd9bc759934b332d433f5.

This contain the fix for go 1.16 that we backported, so we can use
upstream repository again

For now this still have to be enabled with a config like this in a .conf
file in srv/telegraf/extra-config/  :

    [[inputs.execd]]
      name_override = "slapos"
      # this needs sudo when using can not access supervisor socket, like when being installed in root slapos
      command = ["/usr/bin/sudo", "$SOFTWARE_DIR/go.work/bin/telegraf-input-slapos", "-config", "/path/to/slapos.conf"]

/path/to/slapos.conf would contain something like this:

    [[inputs.slapos]]
      ## Folder where partitions are located
      instance_root = "/srv/slapgrid/"

      ## filepath.Glob pattern to look for recursive instances
      recursive_instance_glob_pattern = "*/srv/runner/inst*/"

      ## Path of supervisor socket, relative to instance root
      socket_name = "sv.sock"

This version has a new sql input, that can be used to get metrics from sql
queries.

To run the benchmark locally:

 - Install software/slapos-sr-testing and create an instance
 - Source the shell script published by the instance to setup
   environment variables
 - execute `python_for_test -m unittest -v  test.benchmarks`

This promise was generally used to check the connection quality to the
backend with ICMP, by using ping. This is currently obsoleted by health-check
and other parameters in that group.

The parameter to setup the promise is provided by shared instance requester,
that person has no access to those results, as shared instances can't be
monitored directly, all results are only available to the cluster operator in
contrast to health-check, which results are logged to backend_log, to which
shared instance requester has access.

See merge request nexedi/slapos!1353

1) stack/erp5/zope-versions.cfg
2) stack/slapos.cfg
3) component/ZODB/buildout.cfg.

and remove stack/erp5/ztk-versions.cfg as we need only small part of it.

stack/slapos.cfg should be listed in the begining of software release extends list.

See merge request nexedi/slapos!1348