The allocation_scope of Compute Node is as important, as Compute Node
owner regarding the software management itself.

This is a fixup for 5c938a8d.

this is part of libc

fnmatch.fnmatch signature is path, pattern, we were passing arguments in
reverse order, so effectively we were only checking executable files.
Several .so are executable, so we were checking many .so anyway.

In old Debian 10, after the release of Debian 11, apt update is failing
like this if we don't add the option "--allow-releaseinfo-change":

root@debian:~# apt update
Get:1 http://security.debian.org/debian-security buster/updates InRelease [65.4 kB]
Get:2 http://http.us.debian.org/debian buster InRelease [122 kB]
E: Repository 'http://security.debian.org/debian-security buster/updates InRelease' changed its 'Suite' value from 'stable' to 'oldstable'
N: This must be accepted explicitly before updates for this repository can be applied. See apt-secure(8) manpage for details.

Normally, all compute node are validated.

It allows to have a faster sorting.

Thanks to more specific exceptions the caller is able to understand what
exactly happened during processing, especially useful in case of the caller
uses testcase, and can sneak into real problems during setUpClass.

Some advanced tests setting up clusters need to being able to come in half of
the setup process to do some actions on the cluster in order to have its setup
finished, thus allow to override extracted _setUpClass and separated
waitForInstance, so that those actions can be executed.

The SLAPOS_TEST_SKIP_SOFTWARE_REBUILD changes the default to rebuild always
the software by setting install_all to true.

The SLAPOS_TEST_SKIP_SOFTWARE_CHECK allow to skip software checks, if not
wanted by the developer.

Forgot to update the test.
See nexedi/slapos.core@3ceef26a

Client uses Person_requestComputer, so, allow this script on the compute node token

See merge request !335

   This was done by Thomas, I'm just pushing it into git

    It isn't required to run it that often since everything else will take long, and it perform a too expensive query for run
    it every minute.

   Just save some minor processing time.

  Those are mostly sample files updated for this revision, each project should keep a
  separated dumped files on their on project context.

See merge request nexedi/slapos.core!333

this fixes 86ce8b8e which had an error
the current_output variable was not reset when the current output ended
in \n.

As a result, logged output was sometimes containing some parts of the
program output twice. We could even catch it in unit test, when running
the test multiple times. It also happened on test node with
https://erp5js.nexedi.net/#/test_result_module/20210826-1706D7F50/10

The is trivial, we reset the current output also in this case. The test
suite is extended a bit, with added tests and test adjustments made during
the debugging. With test_exec_large_output_multiline the problem triggers
almost every time.

   This is basically dead code, and untested.

See merge request nexedi/slapos.core!328

Also migrate Computer Partition to Compute Partition

Keep permissions on the computer module to allow computer_module/Base_getComputerToken call

libslap must continue to use the same class names for compatibility.

Changes are made to keep the migration as simple as possible:
  * do not change the catalog table/column names
  * do not change the local_role's names and the related categories
  * keep Computer Consumption portal type
  * keep portal_id's group name
  * keep SLA compatibility
  * keep SlapTool compatibility

SlapTool: restore compatibility with Hosting Subscription method