Fail fast on unknown shell run arguments

[biefan]

Summary

• reject unknown pyrit_shell run arguments instead of logging a warning and continuing
• keep the existing shell error path so invalid run input is shown to the user immediately
• add regression coverage for unknown shell run arguments

Problem

parse_run_arguments() currently logs unknown tokens and keeps parsing. In shell mode that means a typo can be silently ignored while the rest of the command still runs.

For example:

• run demo --bogus 123 --max-retries 2

currently logs warnings for --bogus and 123, but still returns a parsed command with max_retries=2.

That turns invalid CLI input into a silent partial execution instead of a clear parse error.

Testing

• .venv/bin/pytest tests/unit/cli/test_frontend_core.py -q
• .venv/bin/pytest tests/unit/cli/test_pyrit_shell.py -q -k parse_error

[rlundeen2]

I didn't see where this is in code, but hoping this doesn't terminate the process, just re-type. If so that's ideal :)

[rlundeen2]

Fixed it in another pr because we had to consolidate another fix. Closing this one although this is independently a good fix.

See: #1489