tmpfs and swap are correctly setup on ganeti and hetzner cloud, so we
don't need them in the common trunk.

This is mostly so that we know how to design the previous installers.

 * nagios goes first so we get less warnings during decom
 * backups are almost last
 * inventory is last, except for unracking
 * dnswl entry is earlier

I am aware this will break references from checklist in trac tickets,
but i think that's a less evil. Besides we can still refer to those
through the history as needed.

badblocks is slower than nwipe, as it always rereads sector. We can
turn that feature off in nwipe (--verify=off) and have better control
over the patterns written.

We pick the "random" pattern because we don't really trust the US DoD
or Canada federal police with our data.

We need a better way than this...

Modify documentation to add a new ganeti instance with information about where to add a reverse dns entry in hetzner

The rationale here is we remove one more step in the installer:
instead of going back and forth between the master and client, we run
one command on both, simultaneously, and only transfer the SHA256
checksum between the two.

This could eventually be automated in such a way that we can do this
unattended if we have SSH access on both machines.

The rationale is that Trac has all the tickets and is a better place
to follow those ticket's statuses. We're going to try to use the
tickets for february and march, and if that goes well, turn all
actionable text entries into tickets and just use trac for tracking.

required merging ud-ldap and ud-ldap-cgi from Debian