fix: clarify interpretation boundary across viewing, provenance, and prompts

[iqdoctor]

Background

This PR addresses the architectural clarification proposed in #37 and is made explicitly on the basis of the article:
https://ailev.livejournal.com/1795114.html

The article names an open gap around:

• interpretation as conservative reading,
• explanation / justification surfaces,
• interpretation as hypothesis generation.

Problem addressed

Closes #37.

The current spec already has the neighboring owners (A.6.3, A.10, B.5.2.0 / B.5.2), but their boundary remains implicit enough to invite drift between conservative viewing, provenance-bound rendering, and abductive interpretation.

Minimal repair

• add one boundary note to A.6.3 for conservative re-textualization
• add one provenance note to A.10 for faithful rendering vs post-hoc narrative
• strengthen the B.5.2.0 boundary rule so prompts are not confused with conservative renderings

Out of scope

• no new pattern family is introduced here
• no claim that canonical naming / final placement are fully settled
• no changes to the existing G-related PRs (fix: normalize USM checklist language so scope is never typed as characteristic #32, fix: align B.3 G-composition with USM path semantics #34, fix: normalize KD-CAL and Kind-CAL teaching layer for G as scope object #36)