Improve hardware telemetry error reporting by preserving canonical error codes.

Previously, low-level telemetry errors were collapsed into generic unavailable states in the middleware cache, swallowing precise diagnostic details.

This change:
- Updates the telemetry accessor base to cache and expose the precise canonical status returned by hardware read operations.
- Updates the polling sensor interface to map and propagate these granular status codes (such as driver read errors or data corruption) directly to the BMC state tables.
- Ensures downstream interfaces (like Redfish) receive specific, machine-readable symptoms instead of generic failure flags.

Testing:
- Added unit tests verifying telemetry state transitions for valid data, driver read issues, and data integrity failures.
Google-Bug-Id:511201651
PiperOrigin-RevId: 968842688
Change-Id: Ie38f1a4ca6431e436eae5bdb009fdd62b4eb99ba