Confine the auto-resume durability fallback to auto-resume

[Naeemkh]

Summary

• _resolve_dcp_load_dir's durability fallback could never fire. explicit tested whether a path argument was passed, but auto-resume always passes the path it resolved, so every production call took the early return and the fallback below was dead code. Now it tests checkpoint.load_path, i.e. whether the user named a checkpoint.
• Effect: an auto-resume onto a checkpoint whose DCP shards were never made durable now degrades to the last good step instead of dying in dcp.load with metadata is None. An explicit load_path is still honored as-is and still fails loudly.
• Reachable on the 9 shipped configs using async_with_pinned_mem.

Testing

• uv run ruff check kempnerforge/ tests/ passes
• uv run ruff format --check kempnerforge/ tests/ scripts/ passes
• uv run pyright kempnerforge/ passes (0 errors)
• uv run pytest tests/unit/ -v --timeout=60 passes
• If distributed code changed: uv run torchrun --nproc_per_node=4 -m pytest tests/distributed/ -v
• If training loop / parallelism / optimizers changed: uv run pytest tests/e2e/ --e2e -v

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
kempnerforge/checkpoint/manager.py	90.63% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.