Drop two ways contract for machine's failure domain

Since 0.3.0 we are supporting a two ways contract for machine's failure domain, which is really confusing.

Most specifically, we are supporting a deprecated reverse process where InfraMachine was authoritative WRT to failure domain placement (this was implemented for allowing a transparent transition from when there was no failure domain support in Cluster API)

This issue is about creating awareness on the fact that we are going to drop the reverse process entirely with v1beta2.

I also already added an heads up warning to the infra machine contract page in https://github.com/kubernetes-sigs/cluster-api/pull/11223

Rif https://github.com/kubernetes-sigs/cluster-api/blame/274d7e21e1ac2eb89b3f5986f6a09ff9db4eeacc/internal/controllers/machine/machine_controller_phases.go#L324-L333

/kind cleanup /priority important-soon /triage accepted /help