Eliminate requeueAfter Usage for Scalability

[zhanggbj]

/kind feature

Describe the solution you'd like
[A clear and concise description of what you want to happen.]
We currently use requeueAfter in several controller reconciliation loops. While simple, this mechanism may lead to scalability issues.

This issue tracks the work to audit all usage of requeueAfter and replace it with efficient, event-driven watches where possible.

TBD for items to audit.

• VMOMachine Reconciliation.

We should investigate creating watches from objects like VirtualMachine to the owning VirtualMachineGroup or VirtualMachine to ensure that VMG updates are propagated immediately upon a relevant change to a VM, rather than relying on a requeueAfter interval.

Anything else you would like to add:
[Miscellaneous information that will assist in solving the issue.]
Discussion raised from #3655 (comment)

Environment:

• Cluster-api-provider-vsphere version:
• Kubernetes version: (use kubectl version):
• OS (e.g. from /etc/os-release):

[zhanggbj]

CC @fabriziopandini @sbueringer