[live-migration] adds the snapshot proto for the shim state

[rawahars]

Summary

Introduces internal/controller/migration/snapshot, a new package that defines the wire format used to hand off a running LCOW sandbox between shims during live migration.

What's in this change

• snapshot.proto — new protobuf schema describing a self-contained, point-in-time snapshot of an LCOW sandbox: the utility VM, its device sub-controllers (SCSI, VPCI, Plan9), per-pod network state, and the containers and processes running in the guest. The schema is versioned so that incompatible producers and consumers fail fast instead of silently corrupting state.
• snapshot.pb.go — generated Go bindings (produced via protobuild; see Protobuild.toml).
• doc.go — package overview describing the migration use case, the controller-oriented layout, and the source/destination caller flow.

Design notes

• Controller-oriented layout: The schema mirrors the shim's existing controller boundaries: each controller on the source serializes its own bookkeeping into a dedicated section, and the corresponding controller on the destination rebuilds its in-memory state from that section. Adding new migratable state in the future is a localized change to the controller that owns it.
• Functional hierarchy at the top of the .proto. A short ASCII tree summarizes the snapshot rooted at SandboxSnapshot (VM + sub-controllers, and per-pod network/containers/processes), plus a cross-reference table showing which string IDs in ContainerState / LayerReservation index back into the VM-level resource maps. This lets a new reader grasp the shape of a snapshot without scrolling through every message.
• Schema versioning. SandboxSnapshot.schema_version is bumped whenever field semantics change; destinations must reject unknown versions.
• Lifecycle enums. Each Stage enum mirrors the corresponding Go State iota in the controller it represents, with matching zero values so default-constructed messages map to the controllers' initial states.