feat(pydantic-ai): Support span streaming

[alexander-alderman-webb]

Description

Use sentry_sdk.traces.start_span and replace Span.set_data() with StreamedSpan.set_attribute() when in span streaming mode.

Parametrize tests on the trace lifecycle option.

Issues

Closes#6046

Reminders

• Please add tests to validate your changes, and lint your code using tox -e linters.
• Add GH Issue ID & Linear ID (if applicable)
• PR title should use conventional commit style (feat:, fix:, ref:, meta:)
• For external contributors: CONTRIBUTING.md, Sentry SDK development docs, Discord community

[github-actions]

Codecov Results 📊

✅ 282 passed | Total: 282 | Pass Rate: 100% | Execution Time: 43.34s

All tests are passing successfully.

❌ Patch coverage is 0.00%. Project has 14753 uncovered lines.

Files with missing lines (6)

File	Patch %	Lines
ai_client.py	0.00%	⚠️ 156 Missing
utils.py	0.00%	⚠️ 107 Missing
agent_run.py	0.00%	⚠️ 86 Missing
invoke_agent.py	0.00%	⚠️ 82 Missing
execute_tool.py	0.00%	⚠️ 31 Missing
utils.py	0.00%	⚠️ 29 Missing

---

Generated by Codecov Action

[sentry-warden]

update_invoke_agent_span silently drops GEN_AI_RESPONSE_MODEL in span streaming mode

In sentry_sdk/integrations/pydantic_ai/spans/invoke_agent.py, update_invoke_agent_span calls span.set_data() unconditionally, but StreamedSpan only has set_attribute() — the resulting AttributeError is swallowed by the surrounding except Exception, so the response model name is silently never set when span streaming is enabled.

Evidence

• StreamedSpan (defined in sentry_sdk/traces.py:227) has set_attribute/set_attributes but no set_data method (confirmed via grep).
• update_invoke_agent_span calls span.set_data(SPANDATA.GEN_AI_RESPONSE_MODEL, response.model_name) at invoke_agent.py:168 without an isinstance(span, StreamedSpan) guard.
• The call is wrapped in try/except Exception, so the AttributeError is silently eaten.
• ai_client_span and execute_tool_span in the same PR both correctly use isinstance(span, StreamedSpan) checks (e.g., execute_tool.py:update_execute_tool_span).
• The new parametrized tests added in the hunk check gen_ai.chat and gen_ai.execute_tool spans but not the invoke_agent span's GEN_AI_RESPONSE_MODEL attribute, so this regression goes undetected.

_{Identified by Warden code-review}

[sentrivana]

Looks good, small improvement suggestions in tests