Adding HostExec for stages and RestartPolicy

[steiler]

Adding the RestartPolicy as a knob, also cleaning up the docker runtime from referencing specific node kinds.
Further implements host-exec per stage, which executes commands in the host namespace as part of the stages of a specific node.
Tackling the two topics that surfaced in #2233

name: linuxfrr

topology:
  kinds:
    linux:
      image: alpine
  nodes:
    frr1:
      kind: linux
      stages:
        configure:
          host-exec:
            on-enter:
              - docker update --restart=no clab-linuxfrr-frr1
    frr2:
      kind: linux
    frr3:
      kind: linux
      restart-policy: no

The json schema update is missing so far, as well as docs update.

[hellt]

thanks @steiler
I have added docs/schema/tests and did one slight modification by putting the host execs target ahead of the container exec.

The idea is that host execs are typically preparation commands, so they would benefit form running first, and then the container stage execs would follow

[steiler]

The idea is that host execs are typically preparation commands, so they would benefit form running first, and then the container stage execs would follow

I've also thought about orders. The other option would be to reorg that part and make it more like

  nodes:
    frr1:
      kind: linux
      stages:
        configure:
          exec:
            on-enter:
              - command: <command>
                target: <host/container>

That would allow to run them in any order ...

[codecov]

Codecov Report

Attention: Patch coverage is 70.51282% with 23 lines in your changes missing coverage. Please review.

Project coverage is 52.12%. Comparing base (2c249b2) to head (5ea2aa0).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
clab/dependency_manager/dependency_node.go	81.25%	5 Missing and 1 partial ⚠️
runtime/docker/docker.go	50.00%	6 Missing ⚠️
types/stages.go	66.66%	4 Missing and 2 partials ⚠️
types/topology.go	42.85%	2 Missing and 2 partials ⚠️
nodes/host/host.go	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2238      +/-   ##
==========================================
+ Coverage   51.94%   52.12%   +0.18%     
==========================================
  Files         172      172              
  Lines       12658    12714      +56     
==========================================
+ Hits         6575     6627      +52     
+ Misses       5175     5174       -1     
- Partials      908      913       +5     

Files with missing lines	Coverage Δ
clab/config.go	68.62% <100.00%> (+0.10%)	⬆️
nodes/linux/linux.go	76.08% <100.00%> (+1.08%)	⬆️
types/node_definition.go	94.17% <100.00%> (+0.12%)	⬆️
types/types.go	60.75% <ø> (ø)
nodes/host/host.go	34.84% <50.00%> (+17.66%)	⬆️
types/topology.go	71.96% <42.85%> (-0.65%)	⬇️
clab/dependency_manager/dependency_node.go	87.65% <81.25%> (+1.08%)	⬆️
runtime/docker/docker.go	71.54% <50.00%> (-0.53%)	⬇️
types/stages.go	69.50% <66.66%> (-1.60%)	⬇️

... and 2 files with indirect coverage changes

[hellt]

@steiler

The idea is that host execs are typically preparation commands, so they would benefit form running first, and then the container stage execs would follow

I've also thought about orders. The other option would be to reorg that part and make it more like

  nodes:
    frr1:
      kind: linux
      stages:
        configure:
          exec:
            on-enter:
              - command: <command>
                target: <host/container>

That would allow to run them in any order ...

This is way better, I agree
I merged the current version to save the current state, but if you know how to implement this - it would be much cleaner