feat: add FDv1AdapterSynchronizer wrapping IDataSynchronizer as IFDv2Synchronizer

[beekld]

Summary

Stacked on #539. Adds an adapter that wraps any FDv1 IDataSynchronizer (e.g. the existing StreamingDataSource) and presents it as an IFDv2Synchronizer. FDv1 Init/Upsert callbacks delivered through an internal IDestination are translated into FDv2SourceResult::ChangeSet results (kFull for Init, kPartial for Upsert), with empty selectors and fdv1_fallback=false.

The orchestrator picks this up after SwitchToFDv1Fallback() once a factory whose IsFDv1Fallback() returns true is configured. Wiring the adapter into the public config builders is a separate concern, deferred until the Configuration API work.

Test plan

• 8 unit tests cover: lazy start, no-restart on second Next, Init→kFull, Upsert→kPartial (both overloads), pending-Next-resolved-as-Shutdown-on-close, ShutdownAsync called once on close (and not at all if never started), Next-after-Close
• Full server suite green (486 tests)

---

Note

Medium Risk
Touches core FDv2 data-sync orchestration semantics (result types, close/shutdown, status→error mapping); risk is mitigated by isolation behind IFDv2Synchronizer and broad unit tests, but incorrect adapter behavior could affect fallback flag delivery once wired.

Overview
Adds FDv1AdapterSynchronizer, which implements IFDv2Synchronizer by wrapping an FDv1 IDataSynchronizer built via an injectable SourceBuilder. The first Next() lazily calls StartAsync on the wrapped source with an internal IDestination that turns FDv1 Init/Upsert into FDv2SourceResult::ChangeSet (kFull / kPartial, empty selector, no fdv1_fallback directive).

Lifecycle and async behavior: a thread-safe queue/promise State serializes results to outstanding Next() calls; Close() races with pending Next() via WhenAny and resolves waiters with Shutdown, shutting down the FDv1 source only if it was started. FDv1 DataSourceStatus errors map to Interrupted (initializing/interrupted) or TerminalError (kOff).

The new sources are registered in CMakeLists.txt. Unit tests exercise start-once semantics, conversion, FIFO queuing, status mapping, and close paths. Public config/orchestrator factory wiring is not in this diff (intended for a follow-up).

^{Reviewed by Cursor Bugbot for commit 2f68933. Bugbot is set up for automated code reviews on this repo. Configure here.}

[kinyoklion]

Is this ultimately going to be the correct pointer type? The data sources use shared_from_this and weak_from_this?

[beekld]

Fixed. And I filed a task to add a contract test that would've caught this.

[kinyoklion]

Do we need to be concerned with error only status changes? Sometimes we update the error but leave the state the same. So it wouldn't affect the flow of fallback/recovery, but the granularity of the reported status.

[beekld]

Fixed.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 2f68933. Configure here.}