Add evidence-aware Sampled Values profile infrastructure

[masarray]

Purpose

Implement the standards-neutral profile foundation required before adding verified named Sampled Values profiles.

Engine changes

• Add transport-independent frame observations and observation-window accumulation.
• Calculate stable fields, measured frame/sample rates, and observed sample-counter wrap.
• Add evidence-aware profile definitions with sampling, packing, payload, dataset, tolerance, and mandatory source metadata.
• Add an explainable weighted detector with match, conflict, unknown, score, and confidence output.
• Require sufficient evaluated evidence plus dataset and sampling matches before a profile may reach Confirmed.
• Add strict and compatible configuration-versus-wire comparison without blocking receive-side capture or decoding.
• Keep the built-in catalog limited to a generic SCL-driven Layer-2 fallback; no unverified named-profile constants are introduced.

Tests and coverage

• Observation rate, stable-field, and counter-wrap behavior.
• Exact, conflicting, sparse, and insufficient-evidence profile detection.
• Evidence explanations and generic-catalog boundaries.
• Strict versus compatible mismatch behavior.
• Profile-definition source, sampling, and dataset validation.
• 74 deterministic tests pass.
• Whole-engine coverage: 13.35% across 16,312 lines; enforced floor raised to 13%.
• Protocol-core coverage: 70.47% across 2,120 lines; enforced floor raised to 70%.

Public repository language

• Replace proprietary comparison language with neutral engineering outcomes.
• Add an automated public terminology-neutrality check to CI and Pages deployment.
• Retain a validation report artifact for review.
• Document the profile infrastructure, output contract, evidence intake, and acceptance boundary.

Validation completed

• Public terminology validation: success.
• Public-site build and recursive validation: success.
• Locked restore: success.
• Publisher and Subscriber warning-free builds: success.
• Unit tests and raised coverage gates: success.
• Dependency vulnerability reports: success.
• Full-solution CodeQL: success.
• Portable Publisher and Subscriber publication: success.
• Installer compilation: success.
• CycloneDX SBOM validation: success.
• Silent install/uninstall smoke tests: success.
• SHA-256 checksum generation: success.
• Public release job on the PR: correctly skipped.

Not included

• No new public profile-support claim.
• No frozen 9-2LE or IEC 61869-9 numeric definition.
• No WPF or live-capture integration yet.
• No public release.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e6b0823c46

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve exact svID and dataset identifiers

When strict comparison checks svID and dataset reference through CompareText, it also uses the MAC-style normalization that strips -/: and compares case-insensitively. In strict preflight this makes different bindings such as MU-01 vs mu01 or LLN0$Phs-A vs LLN0$PhsA look like an exact match, even though those identifier strings are not MAC addresses and should remain significant. Apply separator-insensitive normalization only to the destination MAC field.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Return Conflict when any evaluated fact contradicts

A low-weight contradiction can be hidden as Likely or Possible here because those thresholds run before the final conflictWeight > 0 check. For example, if all high-weight fields match but nominal frequency conflicts, conflictWeight is less than matchedWeight and the score stays above 70, so the detector reports Likely despite contradictory evidence; the new output contract says to use Conflict when evaluated facts contradict a definition, and consumers sorting/displaying by confidence can otherwise present a conflicting profile as acceptable.

Useful? React with 👍 / 👎.