perf(tracing): span queue linger + per-loop httpx keepalive #1542

Job	Run time
find-tutorials	6s
test-00_sync/020_streaming	2m 33s
test-00_sync/000_hello_acp	1m 58s
test-10_async/00_base/100_langgraph	2m 27s
test-10_async/00_base/030_tracing	2m 4s
test-10_async/00_base/080_batch_events	3m 11s
test-10_async/00_base/000_hello_acp	2m 2s
test-10_async/10_temporal/000_hello_acp	2m 4s
test-10_async/00_base/110_pydantic_ai	2m 9s
test-10_async/10_temporal/030_custom_activities	1m 55s
test-10_async/00_base/010_multiturn	2m 12s
test-00_sync/030_langgraph	3m 4s
test-10_async/00_base/020_streaming	2m 11s
test-10_async/10_temporal/070_open_ai_agents_sdk_tools	2m 4s
test-10_async/10_temporal/050_agent_chat_guardrails	1m 54s
test-10_async/00_base/040_other_sdks	3m 47s
test-10_async/10_temporal/110_pydantic_ai	2m 11s
test-10_async/10_temporal/010_agent_chat	3m 6s
test-10_async/10_temporal/090_claude_agents_sdk_mvp	1m 51s
test-10_async/10_temporal/060_open_ai_agents_sdk_hello_world	2m 4s
test-00_sync/010_multiturn	2m 48s
test-10_async/10_temporal/020_state_machine	2m 29s
test-10_async/10_temporal/080_open_ai_agents_sdk_human_in_the_loop	2m 5s
test-00_sync/040_pydantic_ai	2m 5s
Test Summary	7s
	54m 27s

Provide feedback