lag based on lag.members syntax

[jbemmel]

A variation on the earlier PR:

links:
- lag.members: [ r1-r2, r1-r2 ]

This forces a check for the lag module when using this functionality, it is more concise

Replaces #1337

[ipspace]

You keep adding stuff, so I put the PR back into draft. Please let me know when you're done.

[ipspace]

@jbemmel Is this final and ready for review? Thank you!

[ipspace]

Found an interesting case that crashes netlab: define LAG link group without enabling the LAG module, and the link transformation crashes:

provider: clab

nodes:
  r1:
  r2:

links:
- group: lg1
  type: lag
  members: [ r1-r2, r1-r2 ]

I think this should be fixed. Alternatively, I could merge the PR and then rework it to use lag.members attribute which would force LAG module to be used. It would also get rid of the group link code.

FWIW, if I use...

links:
- lag.members: [ r1-r2, r1-r2 ]

... then netlab complains about missing module before it complains about the lack of nodes on the link, so it looks like we could do the transformation in a module (i.e. it would not require a plugin)

How would you like to proceed?

[jbemmel]

How would you like to proceed?

Thanks for reviewing, I will rework it to use lag.members instead (implying link.type='lag')

[ipspace]

Wow, this was super-fast. Will review it tomorrow, it's getting late over here.

[ipspace]

Great job, just a ton of minor suggestions ;)