Snapshot/restore tooling for Restate environments

[pcholakov]

Requirements:

• customers must be able to capture the complete state of the journal (deployed services, in-flight invocation metadata, persisted KV data) at a point in time into a backup/snapshot
• customers must be able to restore such snapshots into a blank new environment and fully re-create the original environment
• the data format should be open and easy to interpret by standard tools
• snapshots should be compatible across consecutive major releases even in cases where breaking materialization changes prevent a simple online version upgrade

Open questions:

• are there missing features in Restate to force the environment to quiesce before snapshotting? how should we ensure that operators can perform a full snapshot that captures all ongoing service executions prior to a snapshot/restore migration?