refactor(snapshot): add manifest-based snapshotctl flow and shared workload builders

[galletas1712]

Overview

This PR turns snapshot checkpoint/restore into a snapshot-owned protocol layer that is shared by all three entrypoints:

• explicit DynamoCheckpoint CRs
• DynamoGraphDeployment checkpoint restore / mode: Auto
• standalone snapshotctl

The operator side is now much thinner. It keeps CR-specific reconciliation, identity resolution, and status projection, while deploy/snapshot owns the shared checkpoint protocol: labels/annotations, artifact version handling, job naming, lease-aware job observation, restore pod shaping, and snapshot-agent-backed storage discovery.

This also removes operator-owned checkpoint storage configuration. Snapshot storage is now owned by the snapshot chart and discovered from the namespace-local snapshot-agent DaemonSet, so the operator and snapshotctl consume the same source of truth.

High-Level Changes

• add snapshotctl checkpoint and snapshotctl restore for manifest-based checkpoint and restore without the operator
• extract the shared snapshot logic into deploy/snapshot/protocol and reuse it from both snapshotctl and the operator
• refactor DynamoCheckpoint reconciliation so the controller is mostly a thin CR wrapper around shared job/state logic
• move checkpoint storage ownership fully into the snapshot module/chart and remove the redundant operator storage config surface
• rewrite the snapshot docs so the primary flow is explicit DynamoCheckpoint, then DGD auto mode, then lower-level snapshotctl

Reviewer Start Points

• deploy/snapshot/protocol/common.go
• deploy/snapshot/protocol/checkpoint.go
• deploy/snapshot/protocol/restore.go
• deploy/snapshot/cmd/snapshotctl/
• deploy/operator/internal/controller/dynamocheckpoint_controller.go
• deploy/operator/internal/checkpointjob/job.go
• docs/kubernetes/snapshot.md

Smaller Changes

• move the snapshot implementation packages from deploy/snapshot/pkg/... to deploy/snapshot/internal/...
• standardize the snapshot agent selector on app.kubernetes.io/component=snapshot-agent
• add snapshot-owned PVC storage discovery from the snapshot-agent DaemonSet and use it from both operator restore prep and snapshotctl
• add apps/daemonsets read RBAC where the operator now performs snapshot-owned storage discovery
• keep storage.type in the snapshot chart for future expansion, but fail fast for non-pvc values today
• remove checkpoint storage fields, defaults, validation, and Helm values from the operator config
• update DynamoCheckpoint status/docs/generated CRDs to the current phase/hash/job/message model
• simplify the operator Docker/Make/CI wiring by using a named BuildKit context for deploy/snapshot
• delete the duplicate snapshot chart README and consolidate user docs into docs/kubernetes/snapshot.md
• clean up the sample DynamoCheckpoint manifest and snapshot troubleshooting guidance
• use the first container consistently in the shared snapshot checkpoint/restore paths, and document snapshotctl as a single-container flow

Summary by CodeRabbit

• New Features

  • Added snapshotctl command-line tool for manual checkpoint and restore operations.

• Refactor

  • Removed checkpoint storage backend configuration from operator settings; checkpoint storage is now snapshot-agent managed.
  • Restructured snapshot protocol to centralize checkpoint and restore workflows.
  • Reorganized snapshot package structure for better separation of concerns.

• Documentation

  • Updated Helm chart documentation and Kubernetes API reference to reflect storage configuration removal.
  • Revised snapshot workflow guide with new quick-start using DynamoCheckpoint resources and auto-mode checkpointing.

[dillon-cullinan]

Approving changes under .github/workflows