🌱 Remove usage of FailureReason and FailureMessage (baremetal)

[guettli]

Baremetal PR

Deprecated: This field is deprecated and is
going to be removed when support for v1beta1
will be dropped. Please see
https://github.com/kubernetes-sigs/cluster-api/blob/main/docs/proposals/20240916-improve-status-in-CAPI-resources.md
for more details.

Related HCloud PR 1717

Here’s a shorter, clean PR description:

---

Summary

This PR removes all usage of the deprecated FailureReason and FailureMessage fields from the bare-metal (BM) provider. Status handling is now fully aligned with the newer CAPI conditions-based model.

Why

Upstream CAPI is deprecating these fields. Removing them keeps our provider consistent with the updated status API and prevents reliance on legacy status patterns.

Changes

• BareMetalMachine.SetFailure() --> s.scope.SetErrorAndRemediate(). This sets the clusterv1.RemediateMachineAnnotation on the capiMachine.
• During remediation the controller checks if the remediation was started by SetErrorAndRemediate(). If "yes", then no reboot gets done. The remediation stops, so that the machine gets deleted.
• Removed all references to FailureReason and FailureMessage in BM controllers and API types.
• Updated reconciliation logic to use standard Conditions exclusively.
• Adjusted tests to match the new status behavior.
• Test code was adapted to no longer use a fixed bmMachineName. Capi-machine and hbmm are exptected to have the same name. This was adapted, so test data structures match real structures.
• The bm remediation does not take the Namespace from the InfrastructureRef. Cross namespace references are not allowed, because user1 of namespace1 must not reference namespace2.
• Improved logging of unusual cases (conflictError, hbmm is deleting).
• Duplicate code was removed and getAssociatedHost() was refactored to getAassociatedHostAndPatchHelper()
• Remediation code expects that the capi-machine name and hbmm name are equal. Test code was adapted accordingly.