Incorporate SLURM Dashboard calibration feedback

[Charlie-Wang-03]

Motivation

Calibration of slurm-dashboard exposed three reusable distinctions with cross-project value.

Change

• clarify that local or loopback execution does not itself define the real trust boundary;
• require write-capable tools to state their mutation authority explicitly even when most of the product is observational;
• distinguish mocked/simulated validation from real-environment evidence when the environment materially affects correctness.

These refine existing Preferences and Project Conventions without adding a new principle or governance layer.

Validation

Run python3 scripts/validate_repo.py.