[CI] add timeout-minutes: 120 to all pr-test workflow jobs (vllm-project#13171)

## Summary

- Add `timeout-minutes: 120` to every job in the PR test workflow and
all referenced reusable workflows
- Prevents jobs from hanging indefinitely when runners lose connectivity
to GitHub Results Service (as observed with stuck `main2main` jobs)

## Affected files

| File | Jobs updated |
|------|-------------|
| `pr_test.yaml` | `pre-commit`, `select-tests`,
`recommend-tests-from-coverage`, `ci-gate` |
| `_selected_tests.yaml` | `selected-tests`, `upload-coverage-to-obs` |
| `_selected_tests_upstream.yaml` | `e2e-upstream_pr` |
| `_ensure_csrc_cache.yaml` | `prepare-cache-checks`,
`build-missing-caches` |
| `_build_csrc_cache.yaml` | `build` |
| `_analyze_failure.yaml` | `analyze`, `archive-failure-report-to-obs` |
| `_manual-hitest.yaml` | `run-hitest` |

## Test plan

- [ ] Verify PR CI runs complete normally within the 120-minute window
- [ ] Verify that stuck jobs are automatically cancelled after 120
minutes

🤖 Generated with [Claude Code](https://claude.com/claude-code)

- vLLM version: v0.26.0
- vLLM main:
vllm-project/vllm@d02df74

Author: Goalina

.github/workflows/_selected_tests.yaml

@@ -78,6 +78,7 @@ permissions:
 jobs:
   selected-tests:
     name: ${{ matrix.group.npu_type }}-${{ matrix.group.num_npus }} card-${{ matrix.group.partition && format('(part {0})', matrix.group.partition) || '' }}
+    timeout-minutes: 120
     strategy:
       fail-fast: false
       matrix: