Clarify AgentPass product positioning hierarchy

[dinpd]

What changed

• restores a product-led README category: action authorization and execution assurance for AI agents
• establishes the canonical promise that AgentPass controls consequential actions and produces independently verifiable authorization/execution evidence
• adds a canonical positioning document covering hierarchy, audiences, differentiation, standards posture, non-positioning claims, and messaging guardrails
• aligns the README and action-gate roadmap around runtime control, portable provider trust, and execution/outcome assurance
• makes conformance an ecosystem and proof strategy rather than the project category
• aligns the package description and standards-facing documents
• adds regression tests for the public language, hierarchy, boundaries, and new local links

Why

The prior README led with interoperability and conformance, which made the operational product appear to be primarily a standards-testing utility. Issue #73 clarifies the hierarchy without removing or diminishing the existing conformance, provider, runtime, observability, or assurance work.

Impact

This is a positioning, documentation, package-description, and regression-test change only. It does not change runtime behavior, authorization semantics, APIs, dependencies, permissions, data handling, or deployment behavior.

Validation

• python -m pytest tests/test_positioning_docs.py -q — 5 passed
• python -m pytest -q — 109 passed
• Python compile check — passed
• whitespace/diff check — passed
• scoped local-link checks — passed
• manual secret and scope review — no findings

Dedicated static/security scanners were not installed. python -m pip check reports pre-existing global environment conflicts for aider-chat and open-interpreter; this change modifies no dependencies.

Closes #73