Downgrade process is wrong

[TeddyAndrieux]

Component:

'salt', 'downgrade'

What happened:

Current downgrade process:

- Downgrade workload plane nodes and then control plane nodes (except bootstrap one) <-- (including etcd ....)
- Downgrade etcd cluster <-- (automatically done when downgrading other nodes)
- Downgrade k8s objects
- Downgrade bootstrap stuff (salt-master, repositories, ...)

Several issues:

• if we need ext_pillar informations when downgrading nodes or etcd that use some stuff only available with old salt-master we can't as we downgrade the salt-master at the end
• downgrading of etcd it's done when downgrading node and downgrade of etcd basically do nothing

What was expected:

Being able to use old ext_pillar only available with old salt-master when downgrading

Resolution proposal (optional):

TBD:
Changes the downgrade process order:

- Apply bunch of pre-downgrade k8s states (cleanup no longer used objects, ...)
- Downgrade bootstrap stuff (salt-master, repositories, ...)
- Downgrade workload plane nodes and then control plane nodes (removing `etcd` role from pillar when calling highstate)
- Downgrade etcd cluster
- Downgrade kubernetes objects (NOTE: this one may be moved right after "downgrade of bootstrap stuff", TBD)