fix(ci_visibility): correct format string in retry-bookkeeping warning (#17862)

> [!NOTE]
> CC @gnufede @vitor-de-araujo — flagging the CI App Libraries team since this touches `ddtrace/testing/internal/pytest/plugin.py`. Happy to hand this off if you'd rather take it.

## Summary

Fixes a format-string bug in `RetryReports.make_final_report` that has been corrupting CI logs and (we believe) contributing to spurious test failures whenever Datadog Test Optimization's attempt-to-fix retries hit a bookkeeping mismatch.

## The bug

The `log.warning` at `ddtrace/testing/internal/pytest/plugin.py:968` had two placeholders (`%s` and `%r`) but only one argument (`test`). Every time this branch fires, Python's logging hits a `TypeError` while formatting and emits a `--- Logging error ---` traceback to stderr instead of the intended diagnostic. The mismatch was introduced in [#15651](#15651) ([a95ac0d](a95ac0d52c), 2025-12-17), so we've had no usable signal for the underlying bookkeeping bug for ~4.5 months.

## When this branch fires

The `IndexError` branch is reached when a retry handler reports a `final_status` (e.g. `PASS`) but no individual retry was recorded under that outcome in `reports_by_outcome`. **In the dd-trace-py kafka contrib suite this is happening ~70 times per day across py3.9/3.10/3.13 (steady-state since at least 2026-04-27).**

The `IndexError` itself is caught locally and the function returns a degraded report with `longrepr=None` — that part is fine. The real damage is the `--- Logging error ---` output that appears as part of the test's stderr capture, which gets surfaced in pytest output and (in some flows) interacts badly with attempt-to-fix retries to produce spurious FAILED markers.

## How we noticed

We marked some `test_data_streams_payload_size` parametrizations as Active in Flaky Tests Management after #17762 fixed the bucket-boundary race. They re-quarantined within hours. Investigation showed the test logic was passing on every attempt — the failure came from this format-string bug polluting the protocol's stderr.

## Fix

Two small changes in the warning:

1. Add the missing `outcome` argument so the format string actually formats.
2. Include the outcomes that WERE recorded so the deeper bookkeeping bug becomes diagnosable when this branch fires.

## Out of scope

The deeper logic bug — *why* `reports_by_outcome[outcome]` ends up empty when the retry handler reports an outcome — is left for the CI App Libraries team. With this fix in place, the warning is now actionable and should make root-causing it straightforward.

---

> [!NOTE]
> This fix was developed with the assistance of an LLM (Claude).

Co-authored-by: gnufede <federico.mon@datadoghq.com>

Author: johannbotha

ddtrace/testing/internal/pytest/plugin.py

@@ -956,7 +956,13 @@ def make_final_report(self, test: Test, item: pytest.Item, final_status: TestSta
             longrepr = source_report.longrepr
             wasxfail = getattr(source_report, "wasxfail", None)
         except IndexError:
-            log.warning("Test %s has final outcome %r, but no retry had this outcome; this should never happen", test)
+            log.warning(
+                "Test %s has final outcome %r, but no retry had this outcome; this should never happen. "
+                "Outcomes seen: %s",
+                test,
+                outcome,
+                sorted(self.reports_by_outcome.keys()),
+            )
             longrepr = None
             wasxfail = None
 

releasenotes/notes/ci_visibility-fix-pytest-retry-logging-error-c0eda4595b0d3fa6.yaml

@@ -0,0 +1,4 @@
+---
+fixes:
+  - |
+    CI Visibility: Fixes an issue in the pytest plugin where a malformed log call emitted a ``--- Logging error ---`` traceback to stderr during Attempt to Fix retries, polluting pytest output and contributing to spurious test failures.