Implement snapshot-aware scheduler for optimized VM placement

[andre-j3sus]

Is your feature request related to a problem? Please describe.

Enhance the scheduler by integrating a snapshot-aware placement policy to improve cold start performance. A custom placer can be introduced to rank nodes based on snapshot locality—i.e., whether a node has the snapshot cached locally. This optimization aims to reduce startup latency by preferring nodes that already hold the necessary snapshot data, thereby avoiding remote fetches and reducing network overhead.

Additional context

This follows up on snapshot support introduced in issue #823.