fix: remove unnecessary trace_ctx initialization

The trace_ctx variable is only used inside the `if should_trace_cold_start:`
block, so initializing it to None beforehand is redundant.

Also updated comment to explain WHY we capture trace context before
finishing spans (to ensure ColdStartTracer gets the correct trace_id).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

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

Author: jchrostek-dd

datadog_lambda/wrapper.py

@@ -305,14 +305,13 @@ def _after(self, event, context):
 
             status_code = extract_http_status_code_tag(self.trigger_tags, self.response)
 
-            # Skip creating cold start spans in managed instances mode
-            # In managed instances, the tracer library handles cold start independently
+            # Capture trace context BEFORE finishing spans so ColdStartTracer
+            # creates spans with the correct trace_id.
             should_trace_cold_start = (
                 config.cold_start_tracing
                 and is_new_sandbox()
                 and not is_managed_instances_mode()
             )
-            trace_ctx = None
             if should_trace_cold_start:
                 trace_ctx = tracer.current_trace_context()