fix: Preserve recoverable migration state

[nfebe]

Preserve usable migration archives when remote publication or cleanup fails. Migration checks now use each deployment's environment and actual container metadata.

[sourceant]

no_plan

[sourceant]

Review complete. See the overview comment for a summary.

[sourceant]

Code Review Summary

Backups now record per-component, destination, and cleanup results, adding partial and local_only statuses that keep the local archive usable when remote publication or cleanup fails, plus a retry-publication endpoint to re-upload a retained archive. Container network and access checks now resolve deployments from inspected container metadata, and compose validation loads deployment .env files while cleanup ignores failed backup attempts.

🚀 Key Improvements

• Partial and local-only backup statuses preserve the local archive when remote publication or cleanup fails (internal/backup).
• Per-component, destination, and cleanup results capture individual failure detail (internal/backup/types.go).
• Container network membership is read from the container's own metadata, matching names and IDs (internal/networks/manager.go).