Persist native single-dataset saved SQL through the existing Filter API

[VibeSWE]

The native saved-query library already retains names, SQL, datasets, time ranges and charts locally. Users cannot open that library in another browser.

Add persistence for single-dataset entries through the existing Filter schema and CRUD handlers. Reuse extensible Filter metadata for a validated, versioned native mapping for names, relative/absolute ranges and chart configuration, distinct from the Filter schema version.

Acceptance criteria

• Save, list, open, update and delete entries across browsers for the same authenticated owner/tenant, including after server restart.
• Explicit import of eligible local entries is retry-safe and does not create duplicates; retain local data until persistence succeeds and report partial failures. POST currently generates a fresh server ID, so idempotency needs explicit handling.
• Define duplicate-name and concurrent-edit behavior; another browser must not silently overwrite newer work.
• Enforce owner/tenant isolation and dataset permissions on the server. SQL filters are currently listed to users authorized for the stored query’s datasets, not restricted by owner. Define administrator access explicitly; browser filtering is insufficient.
• Validate metadata while keeping SQL usable when optional chart settings are unsupported.

The model requires one stream_name. Keep multi-dataset entries usable locally with a clear local-only indication until their server contract is designed. Shared libraries and automatic bulk migration are outside this first slice.

Suggested order: second; reuse existing storage rather than adding a new service.