Resolve issues with orphaned VMs when process restarts with in-flight jobs

[Rich7690]

Currently if the process restarts for any reason, the trackedVMs list is reset to zero on startup. This causes issues if you have many jobs in flight by leaving the VMs orphaned on the nodes and it has to be cleaned up manually. This addition adds a orka3 vm list on startup to index all the VMs back into the list so it follows similar orphaned checks.
I wasn't sure if this is the best way to check which VMs belong to the given runner scale set aside from prefix checks on the VM names since the orka API can't necessarily filter any other way.

[ispasov]

As different users have different workflows, lets add this behind an option.
It can be enabled by default, but we should allow people to opt-out from it.

[Rich7690]

As different users have different workflows, lets add this behind an option. It can be enabled by default, but we should allow people to opt-out from it.

Thanks for the review. Added an environment variable to configure the behavior.

[ispasov]

@Rich7690 can you rebase your branch against main so the latest change can be applied.
This would allow us to test this in a scenario where the scaleset is preserved

[Rich7690]

Closing in favor of #66