It should be an action on an instance node

See merge request nexedi/slapos.core!607

This is mainly about error handling, simpler and more useful.

- Always warn when a valid SR can't be loaded, with a message that
  contains the original exception (for example, compared to before
  this commit, it will tell if a JSON file can't be found,
  or where JSON has syntax errors).

- Same as previous point if serialisation type is invalid or missing.
  If the caller needs it to transform parameters (parameters file),
  it will raise with the original exception. Otherwise, it falls back
  on json-in-xml. In some places, such fallback is a change of
  behaviour and I have no opinion about it except that at least
  it's now consistent throughout slapos.core.

- Remove warning about RootSoftwareInstance/default: meaningless
  because contradicted the comment, and useless because the transition
  to 'default' is already complete for SR schemas. There are still 3
  lines of backward compatibility code for the rest of slapos.core.

- Don't read the same file several times. Note however that this
  performance fix is only for the SoftwareReleaseSchema class:
  the caller should be fixed to not instanciate several times with
  the same parameters (from do_request & _requestComputerPartition).

See merge request nexedi/slapos.core!621

See nexedi/slapos.core@01bd2af3

- drop Base_getSupportRequestInProgress
- drop SupportRequest_getLastEvent

  This code isnt needed anymore.

   This is easily doable via portal_templates, so no need to keep ad-hoc scripts here.

  not the agregate.

- when instance url, type and sla are modified
- when a Node may match new SLA conditions
- when partition changes
- when Installation is requested
- when allocation supply is modified
- when Allocation Cell is checked

See merge request nexedi/slapos.core!623

   Automatically destroy and invalidate duplicated software installations, in same state. This is helpful if concurrent requests had created multiple supplies.

  This prevents +2 supplies on the request creates more them one software installation with same url on a compute  node.

  The requestSoftwareRelease don't have proper raise in case 2 supplies are done on the same transaction.

   hateoas url is the website but it requires ssl authentication (like portal_slap), while the slapos_master_web_url is the website the user is accessing (no ssl authentication required).

  It seems that was broken for quite long.

   Owner is not set, but Project
   Update typo on field title

   Remove field that uses internal packing list for display location, since this API was completely dropped.

   If we want to reintroduce it will done differently.

  Configure payable required extra effort, and we should keep default simple.

  So drop entries in tests or upgrades since it is not considered anyway.

  Since the Base_getStatusMonitorUrl  uses try/except, it didn't detect that the file was missing.

  Copied from slapos_jio (which was removed).

   Only Computer enable/disable the monitoring now.

   Monitor scope is per computer bases only, so remove the field since it is not applicable anymore.