[Epic] Tracking issue for implementing in-place update contracts

[alexander-demicev]

This is a tracking issue (epic) for in-place updates.

In-place updates were introduced in CAPI starting with v1.11.x. Since we have now bumped to v1.12.x, we can safely begin implementing this feature. Below is a high-level checklist of what needs to be done:

• [In-place updates] Introduce a feature gate #865
  • Add a feature gate for enabling in-place updates.
• Implement contracts (using kubeadm as a reference)
  • Use the kubeadm implementation in CAPI as the primary reference, as it is currently the best and the only example.
  • Note CAPI maintainers had to refactor parts of the Kubeadm control plane provider to make this work correctly (optimizing the rollout process, ensuring maxSurge is respected, and eliminating race conditions).
  • Be prepared for this task to grow in complexity as similar refactoring may be required.
  • Implement a simple "mock" update extension for CAPRKE2.
  • Use the kubeadm CAPI extension as a reference.
• Write a mock update extension for CAPRKE2 #907
• Add E2E coverage #908
• Update documentation #909

Upstream reference implementation: kubernetes-sigs/cluster-api#12291