Initialize workspace fleets with independently configured backends

[floitsch]

An existing artemis.yaml can now initialize a fresh fleet through fleet init, without an organization lookup or network access. Group operations use its declared state, and pod list uses its configured pod backend.

HTTP bindings select endpoints and opaque scopes independently. Named servers reference credentials in local CLI configuration; lookups require matching server types and URLs. The file backend stores fleet identity, groups, and devices separately, with deterministic output and atomic replacement of individual files.

This is the first workspace integration step, stacked on #1261. Existing fleet formats continue to work. Pod building, provisioning, device workflows, rollout, and migration remain follow-ups. Multi-file fleet mutations are not yet transactional. Usage and configuration details are in docs/workspace.md.

Validation:

• 19 CTest checks passed, covering fleet initialization, groups, references, status, storage, and HTTP pod listing/tagging/upload.
• New offline workspace CLI test passed, including group renames with devices.
• New HTTP integration test passed, covering independent endpoints/scopes, shared connections, anonymous and explicit Supabase sessions, and credential isolation.
• CLI analysis and JSON schema syntax checks passed; existing int.parse deprecation warnings remain.