ci: stop swallowing integration test output on failure

[GuanyiLi-Craig]

Problem

When an integration test fails in CI, the job log shows nothing between ##[endgroup] and ##[error]Process completed with exit code 1 — all the diagnostic output (failed test names, tracebacks, API errors) is discarded. This made the 0.0.36 release failure impossible to diagnose from the logs.

Root cause

Each integration step runs:

set -e
output=$(uv run python tests_integration/.../run_*.py)   # stdout captured into $output
echo "$output"                                            # only printed on the NEXT line
if echo "$output" | grep -q "Failed:"; then ...

The runner prints all per-test detail to stdout, which the command substitution captures into $output. With set -e, a non-zero exit from the runner aborts the step at the assignment line, so echo "$output" never executes and the captured output is thrown away.

Fix

Capture the runner's exit status explicitly, always echo the output, and treat a non-zero exit as failure (in addition to the existing "Failed:" grep — the runner already exit 1s on any failed example, so the exit code is the more reliable signal):

set +e
output=$(uv run python tests_integration/.../run_*.py)
status=$?
set -e
echo "$output"
if [ "$status" -ne 0 ] || echo "$output" | grep -q "Failed:"; then ...

Applied uniformly to all 12 integration-test steps.

Notes

This is a pure CI-observability fix — no test or framework code changes. It does not change which tests run or pass; it only ensures failures are visible in the logs.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.