docs: clarify Nano implementation boundaries

[AetherAI3]

What changed

• Rebuilt the README around Nano's implemented contract: a deterministic strategy DSL and reference runtime that emits intents for a host-owned decision gate.
• Added concise reference documentation for architecture, v0.1 language semantics, and implementation status.
• Reclassified the paper series and experimental modules as design/research material where appropriate, preventing roadmap claims from reading as shipped behavior.
• Corrected misleading build and security claims, documented the source/IR corpus accurately, and aligned package metadata with the implementation.
• Added README-example regression coverage and CI verification that documented strategy assets are included in built wheels.

Why

The prior documentation mixed research vision with the current Python implementation. Generic compiler/runtime terminology, plus unsupported claims such as type checking, content-addressed strategy IR, LLM escalation, and persistent core audit storage, made the project appear much broader than the code.

This PR makes Nano's real value clear: replayable threshold-rule evaluation with an explicit intent/gate boundary.

Impact

New users get a runnable quickstart and accurate boundaries. Contributors get an authoritative documentation hierarchy and a packaged strategy corpus check in CI. Research directions remain available without being mistaken for current API guarantees.

Validation

• python -m compileall -q nano tests
• README Nano example regression test
• README direct-runtime and DecisionGate quickstart execution
• Local Markdown-link, encoding, package-data, and git diff --check validation