/reviewed-on nexedi/slapos.toolbox!12

  In distribution systems, it is possible that master didn't indexed the Software Installation
  when we check, this change will consider we don't know the state of the Software Installation on
  this stage. The code will retry later anyway.

  Update the tests after the fix on the code, as the map now contains all the tests form all the groups

 This bug was preventing to get a single test per group (computer) 
 per time. Causing multiple tests run on the same computer and no
 tests run on others. 

  Implement a watchdog to kill process which are hold on dead lock
  state after a greceful reload.

qmp drive-backup command should exit when qemu has finished to backup image
else image is copied when qemu is still writing and result to a broken backup

If qemu return error in command, raise instead of silent exit with code 0

  This Workarround should be removed on future versions of webrunners

  In case haproxy were not working before.

  Export will fail if backup is run while export, so be more flexible.

  Missing import

   As we don't read entire file, the regex can eventually get
   garbage on the latest line to parse. This commit make it tolerant
   while check.

This is required when the only criterion is on the command-line.
Since we have wrappers to work around the shebang limitation,
the --name option can't be used anymore for Python applications.

  Pin version of setuptools as >=34 don't work

resiliencetest: Verbose log between the interactions to know if the software is working for failing.

  It means there is no extra test to run.

  The full log is available on suite.log at the erp5testnode anyway.

  taskdistribution works on failover mode if master is undefined, so it is not needed to have 2 methods