If the service crash it will inform master by trigger a "bang" call

Learn more by reading: README.update-hash.rst

This repairs buildout profiles on python 2.

To backport the patch on old profiles, a simpler version is to append this to buildout.cfg :

```
[versions]
pyrsistent = 0.16.1
```

See merge request nexedi/slapos!908

some depends egg have special installation methods, we need more than
just listing the egg to install it.

1.0.167 can not longer be installed, 1.0.167.1 is 1.0.167 with a fix
backported.

pyrsistent has a setup_require of pytest-runner. Buildout installs
setup_require from their latest version, regardless of [versions]. When a
new version of pytest-runner was released without python 2 support this
broke installation of software releases on python 2.

This reverts commit c2a7ed05.

See merge request nexedi/slapos!907

`autorun` controls the `slapos-node-auto` service:
  - `running`: start `slapos-node-auto`
  - `stopped`: stop `slapos-node-auto`
  - ( else ) : leave `slapos-node-auto` as it is

See merge request nexedi/slapos!888

slapgrid is the old name of slapos node command.

This reverts commit b3394953.

See merge request nexedi/slapos!904

Jscrawler is a Nodejs app used to crawl a website and build the sitemap xml file.
Sitemaps are downloadable through http server deployed in the SR.

Test has been updated, as it was not really testing parameter given to the
user. Currently the situation is not perfect, but good enough.

Real test would require full monitor URL implicit usage understanding which
is out of scope of this test.

See merge request !903

Then using caucase implies monitor, and it's not required to do so in the
softwares using monitor.

Some old version of plugins have been removed from openvsx when they
changed licence. Uploaders who did not sign the licence had their extensions
removed. This was the case for some builtin vscode extensions uploaded by
openvsx script.

Adding monitor stack allows to use the monitor, then passing the profile and
extending it allows to correctly deploy.

Since haproxy checks the test server for availability and does it
asynchronously, the test server has to reply ASAP for given request, but when
one request blocks, then whole server is blocked, so just simply switch it to
threading model.

It's not the best, but good enough for test usage.

backend-active-check is really long and technical name, whereas health-check
is well known description of backend checks.

X-Reply-Status-Code follow usual way to configure responses.