Improvements on finalize allocation function.

[Gerrit91]

Improves the sequence of the finalize-allocation endpoint handler:

• Runs MachineUpdate after the SwitchUpdate
  • SwitchUpdate has a higher chance to run into concurrent modification errors (Failure to report reinstall by metal-hammer results in fresh install with wiped disks #207) and we don't want to leave reinstall mode before the machine was put into the right VRF
  • Also for regular installations we don't want to boot the target OS in the next loop when the machine is not in the right VRF
• Allocation.Succeeded gets set (Machine allocation succeeded flag never gets set #209) preventing metal-core to re-run the hammer after finalizing the allocation, probably also having an impact on gardener-extension-provider-metal firewall reconciliation

Closes #209.

[Gerrit91]

Tried in mini-lab successfully along with the metal-hammer:

metal_api_image_tag: pr-finalize-allocation-improvements
metal_hammer_image_url: https://images.metal-stack.io/metal-hammer/pull-requests/57-wipe-disks/metal-hammer-initrd.img.lz4

[majst01]

We also should add a retry logic to the updateSwitch Logic in a additional PR