fix: shared enum for gen_ai.operation.name values (#6416)

[DawnofGenX]

Summary

Follow-up to the review discussion on #6404 (r3304429118): the gen_ai.operation.name span attribute values were hardcoded string literals scattered across the AI integrations (chat, execute_tool, invoke_agent, ...).

This PR introduces a shared GENAIOPERATION string enum in sentry_sdk/consts.py (following the existing SPANTEMPLATE(str, Enum) precedent) and replaces all 58 static literal usages SDK-wide with enum members:

• sentry_sdk/consts.py — new GENAIOPERATION enum with the 8 values currently emitted by the SDK: CHAT, CREATE_AGENT, EMBEDDINGS, EXECUTE_TOOL, HANDOFF, INVOKE_AGENT, RESPONSES, TEXT_COMPLETION.
• Replaced literals in: anthropic, openai, openai_agents (ai_client / execute_tool / handoff / invoke_agent), pydantic_ai (ai_client / execute_tool / invoke_agent), langchain, langgraph, litellm, mistral, google_genai, and tracing_utils (the @ai_track templates).
• GENAIOPERATION is a str subclass with __str__ returning the value (identical semantics to SPANTEMPLATE), so f"{GENAIOPERATION.CHAT} model" (used by litellm span names) and every serializer path produce the exact same bytes as before.
• The dynamic operation_name in huggingface_hub derives from the OP constant, so it is unchanged by design.

Wire values are byte-identical — this is a pure refactor, no behavior change.

Closes #6416

Wire-compatibility evidence

Ran the pydantic_ai integration suite (which asserts gen_ai.operation.name values on captured envelopes) on unpatched upstream/master and on this branch, same env, same command:

# BEFORE (upstream/master, changes stashed):
$ TESTPATH=tests/integrations/pydantic_ai/test_pydantic_ai.py uv run tox -e py3.14-pydantic_ai-latest
====================== 331 passed, 63 warnings in 23.32s =======================
  py3.14-pydantic_ai-latest: OK (25.96=setup[0.11]+cmd[25.85] seconds)
  congratulations :) (26.43 seconds)

# AFTER (this branch):
$ TESTPATH=tests/integrations/pydantic_ai/test_pydantic_ai.py uv run tox -e py3.14-pydantic_ai-latest
====================== 331 passed, 63 warnings in 23.03s =======================
  py3.14-pydantic_ai-latest: OK (25.75=setup[0.11]+cmd[25.63] seconds)
  congratulations :) (26.23 seconds)

Additionally verified serialization identity directly:

>>> json.dumps({"gen_ai.operation.name": GENAIOPERATION.CHAT}) == json.dumps({"gen_ai.operation.name": "chat"})
True
>>> msgpack.packb({"k": GENAIOPERATION.CHAT})   # b'\x81\xa1k\xa4chat'  — plain "chat" on the wire

Test plan

• New tests/test_consts.py: pins every GENAIOPERATION member name and value to the exact wire strings, plus str()/f-string/json.dumps round-trips — a rename or revalue now fails CI (regression guard requested in Create an enum for possible gen_ai.operation.name values #6416).
• py3.14-common (full core suite incl. new tests): 2252 passed, 174 skipped
• py3.14-pydantic_ai-latest: 331 passed (before/after above)
• py3.14-anthropic-latest: 587 passed
• py3.14-openai_agents-latest: 305 passed
• py3.14-openai-base-latest: 686 passed
• py3.14-langchain-base-latest: 625 passed
• py3.14-langgraph-v0.6.11: 146 passed
• py3.14-litellm-latest: 171 passed
• py3.14-mistral-latest: 88 passed
• py3.14-google_genai-latest: 526 passed, 1 skipped
• uv run ruff check --fix sentry_sdk tests → All checks passed
• uv run ruff format sentry_sdk tests → formatted
• uv run --group typing mypy sentry_sdk → Success: no issues found in 194 source files

[alexander-alderman-webb]

Going to close this because we don't have capacity to review the 5 PRs you've opened.
If you'd like to contribute dedicate your time towards making one good PR. Thanks.