[Multi-provider] outbound event channel sends can block during shutdown

[jonathannorris]

Context

In multiprovider.go, the forwardProviderEvents loop sends events to p.outboundEvents with a direct channel send. If the channel buffer is full, this blocks indefinitely and can stall the forwarding worker, potentially preventing shutdown from completing.

This affects both the ConfigurationChanged pass-through path and the existing state-change forwarding path.

Suggested Fix

Use a select that also listens on workerCtx.Done() so the goroutine can exit cleanly under backpressure or shutdown:

select {
case <-workerCtx.Done():
    return
case p.outboundEvents <- e.Event:
}

Identified during review of #480.

[jonathannorris]

Closing — after analysis, this is not a realistic concern in practice. The outbound channel buffer equals the number of providers (typically 2-3), the consumer goroutine runs a tight read loop, and provider events are infrequent lifecycle events. The buffer would only fill if the SDK's event executor goroutine itself deadlocked while multiple providers simultaneously burst events.