Provisioning: retry dashboard provisioning while the folder API is unavailable

The function isRetriableFolderError checks status.Code(err) directly, but status.Code returns codes.Unknown for errors that are not a gRPC status or are wrapped with fmt.Errorf("...: %w", err). The gRPC status library's FromError helper is the correct way to unwrap the error chain. If the underlying gRPC status error is wrapped (e.g. fmt.Errorf("%w: %w", dashboards.ErrGetOrCreateFolder, status.Error(...))), status.Code(err) will return codes.Unknown and the retry will not trigger, defeating the purpose of the change. The test at line 117 wraps the gRPC error with fmt.Errorf("%w: %w", ...) which happens to work because status.Code does unwrap one level via errors.Unwrap, but deeper wrapping or different wrapping patterns would break.

In provisionDashboardsWithRetry, time.After creates a timer that is not explicitly stopped when the context is cancelled. If the context is cancelled during the backoff, the function returns immediately but the timer remains allocated until it fires (up to 15 seconds). This is a minor resource leak. Consider using time.NewTimer and stopping it on cancellation to free the timer resource promptly.