Unify run dir enumeration for tracer listing and resolve

[AlexanderOnischenko]

Motivation

• Fix disparity between history-based case runs and filesystem scans so tracer listing tools show a consistent, unified view of runs.
• Normalize historical run_dir values (absolute, repo-relative, DATA_DIR-relative, and .runs/... suffixes) to avoid missed matches.
• Auto-detect nested runs roots (e.g. .runs/runs/runs) and preserve history-only entries in listings with clear status flags like missing_on_disk.

Description

• Introduce inventory types and helpers: RunDirInventory, HistoryRunDir, CaseRunListing and functions _load_history_entries, _normalize_history_run_dir, _detect_runs_roots, _iter_fs_run_dirs, and _collect_run_dir_inventory in src/fetchgraph/tracer/resolve.py to merge history and FS sources.
• Add list_case_run_listings to produce listings that include history-only entries and new statuses (missing_on_disk, missing_case_dir, missing_events, tag_mismatch, missed), and make scan_case_runs history-aware when ordering results.
• Update CLI in src/fetchgraph/tracer/cli.py to use the new listing path for --list-matches and to expand --print-resolve output with runs_root_cli, runs_root_effective, history_path and counts (history vs fs) including warnings when history entries are missing on disk.
• Add tests in tests/test_tracer_run_dir_inventory.py covering relative history path normalization, nested runs root discovery, and reporting of history entries missing on disk.

Testing

• Ran pytest -q tests/test_tracer_run_dir_inventory.py and all tests passed (3 passed).
• The new tests validate relative history path normalization, nested runs_root scanning and missing_on_disk reporting and succeeded.

---

Codex Task