One health score, computed once

Before this change, three different parts of BlueCrow could look at the same system and disagree with each other. The dashboard's health score, INGRID's briefing, and the Optimization Advisor's recommendations were each computed by their own independent logic — same telemetry in, sometimes different conclusions out.

That's fixed now. There's a single health-evaluation pipeline, and everything else reads from its result instead of recomputing its own version:

• The dashboard's 0–100 score and rating come from one place.
• INGRID's briefing narrates the same evaluation, not a second one with its own thresholds.
• The Optimization Advisor's recommendations are derived from the same health result, not a separate set of ad-hoc checks.

We also closed a recurring bug pattern along the way: a few places were treating "0" as a real reading instead of "no data available" — most notably an AMD CPU temperature sensor that wasn't being matched by name correctly, so it silently reported 0°C instead of "unavailable." That's fixed for both Intel and AMD hardware now.

None of this changes what BlueCrow can tell you. It changes whether you can trust that two different screens are telling you the same thing.