slapos_cloud: remove constraints on software instance certificate, revoke previous instance certificate when request a new one.

If a valid certificate exists and Node try to request a new certificate, the old certificate is revoked first.

slapos cli: register computer uses new certificate generation style, configure client generate csr with random CN

Store Certificate ID into a new portal type, update getcertificate and revokeCertificate for person, computer, software instance

-  Move all Computer attribute initialization into Computer constructor
-  Rename 'construct' to 'format' because we are in slap.format
-  Introduce Manager base class
-  Implement CGroupManager to take care of cgroup's cpuset

-  Add TUN class
-  Check Interface is a bridge before performing bridgey operations
-  Replace outdated 'tunctl' binary with 'ip tunctl'

  This raise my lead to a significant change on the API, which
  may require to refactor several tests, to this change keep s a
  backward compatible API.

This reverts commit 889711ed

This approach to solve checking of commands by logging is wrong. It can lead to malicious commands being executed so this test is no longer reliable.

We just should not test issued commands - test should use an isolated environment and then inspect the impact.

/reviewed-on !15

Fix parameters order to use request()

/cc @rafael @alain.takoudjou 

/reviewed-on !13

When creating a TAP interface we try to add it to a bridge no mater what kind of the external interface is. Check whether it is a bridge first.

/reviewed-on nexedi/slapos.core!12

slapos/cli/entry.py, slapos/cli/format.py, slapos/format.py: Do not double the log output when "slapos node boot" failed.

Pre-check if the instance is allocated before check if it was allocated on the right place. This problem seems unreproducible on a running instance.