fix(cli): error when --side-csv is named but missing (don't silently stub)

[dchaudhari7177]

Problem

_ensure_side_csv synthesises a header-only stub when the side-effects CSV is absent. That's right when --side-csv is unspecified, but it also fired when a user explicitly passed a path that doesn't exist (e.g. a typo), silently masking the mistake — inconsistent with --dose-csv/--layer-csv, which both error.

Fix

In main, after the dose/layer existence checks:

• --side-csv unspecified → stub as before (empty Side-effects table)
• --side-csv PATH with missing PATH → parser.error("side-effects CSV not found: PATH"), non-zero exit

_ensure_side_csv itself is unchanged, so the stub path behaves exactly as today.

Tests

New test passes --side-csv <tmp>/typo.csv with the real artifact dose/layer CSVs and asserts SystemExit with a non-zero code and the clear message on stderr. The existing render test already pins the unspecified-stub behaviour.

python -X utf8 -m pytest tests/test_cli.py passes (the -X utf8 is only for my Windows console; unrelated tests print ‖).

Fixes #19

[bamdadd]

Merged — thank you, and welcome to steerbench. This is a clean first contribution: it draws the exact line the CLI needed (an explicitly named --side-csv that's missing is a user typo and should fail loudly, while an unspecified one still falls back to the stub), it reuses the existing parser.error style so the message reads consistently with --dose-csv/--layer-csv, and the test pins both the non-zero exit and the message. Nice work closing #19.

If steerbench is useful to you, a star helps others find it and Watch will keep you posted on new issues — and I'd love more contributions whenever you're up for it.