check-ci: multiple improvements

- List bridge jobs alongside regular jobs so failed triggers like
  deploy_to_reliability_env show up in --list-jobs output
- paginated_get reads X-Total-Pages from the first response and fetches
  all remaining pages in parallel instead of sequentially (~3x speedup
  on paginated endpoints)
- get_child_pipelines fetches sibling children in parallel
- Merge success/failure/counter passes over all_jobs into one loop
- api_get now returns headers as a third tuple element
- Remove dead 401/403 re-check in discover_pipeline
- Remove .replace("Z", "+00:00") workaround (Python 3.11+ handles Z)
- Fix any -> Any type annotation
- Bridge/trigger jobs and regular jobs with allow_failure: true are now
  counted in allow_fail= on the status line instead of failed=, and no
  FAILED: line is printed for them. fetch_all_jobs now also fetches
  bridge jobs so their allow_failure flag is visible.
- show pipeline name (from bridge job) and job IDs in --list-jobs output
- merge pipeline/job/bridge discovery into one pass
  (discover_pipelines_and_jobs) eliminating double bridge fetches and
  separate fetch_pipeline_statuses calls
- fix compute_duration to use GitLab's duration field (not created_at
  queue time)
- replace sys.exit() in api_get/discover_pipeline with GitLabError
  exception
- open success.txt/failure.txt once per poll cycle instead of per job
- warn when paginated_get receives a non-list response
- remove unused allow_fail_count

Author: cataphract