cluster_pilots.py shows a traceback for a bad --k and leaves outputs when --truth is invalid #232

Describe the bug
Two behaviours of inference/cluster_pilots.py on the #225 branch. An out-of-range --k ends in a raw ValueError traceback. If --truth is invalid, the group and centre CSVs have already been written when the error is raised, so a failed run leaves half its outputs.

To Reproduce

1. Run generate_synthetic_pilots.py, then compute_features.py on its output.
2. python inference/cluster_pilots.py --k 1 prints a traceback ending ValueError: k must be 2 to 5 for 15 pilots, got 1.
3. Repeat a row in inference/Data/synthetic_pilot_tiers.csv, then run python inference/cluster_pilots.py --k 3 --truth inference/Data/synthetic_pilot_tiers.csv. pilot_clusters.csv and group_centres.csv are written, then ValueError: Pilot(s) repeated in truth: P01.

Expected behavior
A one-line usage error and exit status 2 for a bad --k, as --truth without --k already gives. No output files when the run fails.

Screenshots
Not applicable.

Additional context
Fit and score first, then write; use parser.error for input errors. Needs CLI tests in tests/test_cluster_pilots.py.