Fix trace context leakage for secondary workers

Secondary workers were restoring trace context twice:
1. From invocation.traceparent (correct - from client)
2. From init_input.traceparent (redundant - from primary worker)

The second restore at line 1329 was problematic because:
- The trace_token was never detached, causing context leakage
- It overwrote the active context from invocation.traceparent
- This could cause spans to not appear under the expected trace

Since secondary workers already receive traceparent in the invocation
(which is the same one the client sent), the restore from init_input
is unnecessary. The init_input.traceparent is still stored for potential
use by workers that don't receive traceparent in the invocation.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: rustyconover

vgi/worker.py

@@ -1321,15 +1321,11 @@ def _process_invocation(
                             tracing.VGI_EXECUTION_ID, exec_id.hex()
                         )
 
-                    if (
-                        instance.init_input is not None
-                        and instance.init_input.traceparent is not None
-                    ):
-                        # The traceparent is passed down from the global init.
-                        trace_token = tracing.restore_trace_context(
-                            instance.init_input.traceparent,
-                            instance.init_input.tracestate,
-                        )
+                    # Note: Secondary workers already have trace context restored
+                    # from invocation.traceparent (line 1458 in _run_loop), so we
+                    # don't need to restore from init_input.traceparent here.
+                    # The init_input.traceparent is stored for potential use by
+                    # workers that don't receive traceparent in the invocation.
 
             except (KeyboardInterrupt, SystemExit):
                 raise