chore: clarify duplicate-check coverage for semantic UI repetition

[tunahorse]

Problem

The current duplicate-code gate did not catch a UI renderer path that repeated exception details already present in the user-facing exception message.

This is expected for the current gate because it detects similar source-code blocks, not semantic repetition in rendered output. There is also a confusing mismatch between the duplicate-code threshold documented in pyproject.toml and the stricter threshold used by the pre-push hook.

Expected behavior

The harness and related documentation should make it clear what duplicate checks do and do not cover. Repeated user-facing facts in error rendering should be caught by an appropriate review or validation path before they land.

Definition of Done

• The enforced duplicate-code threshold is easy to identify from the repository's source of truth.
• The harness documentation accurately describes the duplicate-code gate's coverage and limits.
• There is an observable check, test, or review checklist item for avoiding repeated user-facing error details in UI rendering.
• Existing duplicate-code checks still run successfully after the change.