fixes undefined group['node'] if missing from hosts.ini

[ioagel]

• consolidates handling of master taints in group_vars, all.yml

...

k3s_node_exists: "{{ 'true' if groups['node'] | default([]) | length >= 1 else 'false' }}"

...

extra_server_args: >-
  {{ extra_args }}
  {{ '--node-taint node-role.kubernetes.io/master=true:NoSchedule' if k3s_node_exists | bool else '' }}
  --tls-san {{ apiserver_endpoint }}
  --disable servicelb
  --disable traefik

There is no need for users to mess with the roles even for the taint functionality. The addition to extra_server_args avoids duplication in files: roles/k3s/master/tasks/main.yml and roles/k3s/master/templates/k3s.service.j2. For k3s anyway, taints are another extra server arg.
Now even if a user wants to run for example, 3 master and 2 worker nodes with all of them being schedulable, they just have to delete the taint entry in extra_server_args variable.

The new addition in extra_server_args: --tls-san {{ apiserver_endpoint }} is mandatory, because the ip of our virtual ip needs to be in the SANs of the api server certificate, otherwise we cannot access our cluster.

About testing and linting

In the ipv6 molecule test, because we override the extra_server_args, I added the 2 new entries:

• --tls-san {{ apiserver_endpoint }}
• {{ '--node-taint node-role.kubernetes.io/master=true:NoSchedule' if k3s_node_exists | bool else '' }}

The rest of the changes were to make the linter happy, despite my commits had nothing to do with breaking the linter. I wonder if anyone could make any pull request without the linting action failing... ;-)
I think it is fixed even in a hacky way!
By the way I noticed intermittent failures in the metallb wait task in GitHub actions and I had to increase it to 120s. After that I had no failures.