(examples revamp): remove redundant files

[tinalenguyen]

these examples will be removed because they are already present in another example or are already in the docs:

file	reason for removal
./avatar_agents	all documented on respective plugin pages
./other/text-to-speech	docs
background_audio.py	docs
email_example.py	built in via GetEmailTask()
nvidia_test.py	functionally equivalent to basic_agent.py
realtime_video_agent.py	docs
realtime_with_tts.py	docs
web_search.py	opt for provider tools instead
silent_function_call.py	docs for cancel_tool_reply
phonic_realtime_agent.py	docs
multi_agent.py	favor examples and restaurant agent
realtime_turn_detector.py	<!> maybe needs docs
session_close_callback.py	docs
resume_interrupted_agent.py	docs
realtime_load_chat_history.py	docs
annotated_tool_args.py	displayed in full examples
timed_agent_transcript.py	<!> need docs for getting json format timed transcript, docs
speedup_output_audio.py	<!> should be added to docs here
ultravox_realtime_api.py	docs
dynamic_tool_creation.py	docs
zapier_mcp_integration.py	docs
weather_agent.py	dupe of basic_agent.py
structured_output.py	docs
toggle_io.py	docs
speaker_id_multi_speaker.py	docs
raw_function_description.py	docs
inactive_user.py	<!> maybe needs docs
langgraph_agent.py	<!> maybe needs more docs here
long_running_function.py	<!> docs lead straight to the github example, we should add docs on it in Advanced LLM tab
minimal_worker.py	too minimal?
elevenlab_scribe_v2.py	docs
kokoro_tts.py	<!> should we opt for openai.TTS.with_kokoro()?
legacy_api.py	old, dupe of minimal worker
video-publisher.py	docs
translator.py	select STT providers natively support this
flush_llm_node.py	<!> maybe needs docs
text_only.py	docs
transcriber.py	docs
participant_entrypoint.py	<!> maybe needs more docs here
e2ee.py	docs

file	action
realtime_joke_teller.py	<!> document nova sonic specific function tool outputs
push_to_talk.py	<!> should be documented
async_tool_agent.py	<!> should be documented
fast-preresponse.py	<!> should be documented
error_callback.py	<!> should be documented
langfuse_trace.py	<!> should be documented
instructions_per_modality.py	<!> should be documented
restaurant.py	<!> should be documented
tool_search_agent.py	<!> should be documented

[devin-ai-integration]

Devin Review found 1 new potential issue.

🐛 1 issue in files not directly in the diff

🐛 Missing follow_redirects=True in streamable HTTP client causes redirect failures (livekit-agents/livekit/agents/llm/mcp.py:309-312)

The old _create_http_client method (removed in this PR) explicitly set follow_redirects=True when creating the httpx.AsyncClient. The new inline client creation at livekit-agents/livekit/agents/llm/mcp.py:309 omits this parameter, and httpx.AsyncClient defaults to follow_redirects=False. This means any MCP server behind a redirect (common for auth flows, load balancers, CDNs) will fail with the streamable HTTP transport. The old factory was documented as "ported from mcp.shared._httpx_utils.create_mcp_http_client", suggesting redirect support is expected behavior for MCP clients.

View 7 additional findings in Devin Review.