feat: specify an ovn underlay network through MicroOVN

[gabrielmougard]

closes #239

[masnax]

Looks good, hope the refactor wasn't too annoying. I Just have a few nits.

[gabrielmougard]

@masnax it seems to solve the issue. But now the interactive combination tests seem to fail at some point.

[masnax]

@gabrielmougard Needs a rebase.

[gabrielmougard]

@gabrielmougard it looks like the basic tests aren't passing because the test_add_services set of tests don't have an expected answer for the ovn_underlay_network when calling service add.

With the changed behaviour, microcloud init should ask about the underlay network, but running microcloud service add should skip it, so in each test case you need to set and unset the environment variables accordingly.

It would be helpful to also add a test to cover the case where you run microcloud add in the following situation:

• microcloud init was run to set up 2 nodes with MicroOVN uninstalled (so setup for it was skipped)
• MicroOVN was then installed on both systems (but not initialized)
• microovn cluster bootstrap was run on a "new" 3rd node
• microcloud add was called to add this new 3rd node.

In this case, we should verify that MicroCloud offers micro01 and micro02 to add their underlay configuration, but skips micro03 since it was already set up, even though technically micro03 is the one joining the overall MicroCloud cluster.

I added the test case @masnax Tell me what you think about it ;)

[masnax]

The preseed logic seems to be incorrect, but it's a small fix and since you're going away for a couple weeks, I'll merge this and make a follow-up PR. I'll tag you so you can see how the preseed IPs should be determined.

We also need to update the preseed tests for the new keys, and add the new preseed key to the docs as well.

[masnax]

c.state is not populated for preseed since the recipe is precise, we don't want to change any existing system configurations using a preseed file.

[masnax]

The input is not a cidr, so this will fail.