fix(pytest): capture logs for fixture setup failures

[gibiw]

Summary

• When a test fails during fixture setup, pytest reports it as ERROR and the call phase never runs
• The previous condition call.when == "call" meant _attach_logs was never invoked for setup-phase failures
• log.txt, stdout.txt, and stderr.txt were missing from Qase results for these errors
• Extended the condition to also attach logs when call.when == "setup" and report.failed
• Unlike PR fix(pytest): capture logs for fixture setup failures (ERROR results) #470, this avoids duplicate attachments for passing tests by only capturing setup logs on failure

Changes

• qase-pytest/src/qase/pytest/plugin.py — extended log attachment condition to cover setup failures
• qase-pytest/tests/tests_qase_pytest/test_plugin.py — 4 unit tests covering the new behavior via hookwrapper simulation
• examples/single/pytest/tests/test_fixture_setup_failure.py — example tests demonstrating the fix

Test plan

• Unit tests: test_logs_attached_on_setup_failure — logs captured on fixture ERROR
• Unit tests: test_logs_not_attached_on_setup_success — no duplication for passing tests
• Unit tests: test_logs_still_attached_on_call_phase — existing behavior preserved
• Unit tests: test_logs_not_attached_when_capture_logs_disabled — respects config
• Integration: example tests produce correct JSON report with attachments

Closes #470 (alternative implementation)