Research: compare Semantic Projection Networks against multiary lens theory

[cemphlvn]

Summary

Adds an adversarial comparison between the proposed Semantic Projection Network (SPN) framing and prior work on:

• networks of models;
• multidirectional transformations;
• multiary delta lenses;
• composition and consistency restoration.

The note explicitly attempts to reduce SPNs into established bx vocabulary first. It concludes that the graph topology, n:n relations, and transformation composition are largely absorbed by prior art.

The remaining candidate OSIL contribution is narrowed to:

• heterogeneous computational ecosystems;
• property-specific preservation/loss contracts;
• per-hop evidence;
• path-level contract/evidence composition;
• semantic-preservation-constrained path and realization search;
• self-application.

It also decomposes a contracted projection into T (transformation), C (contract), and E (evidence), and proposes candidate composition laws as research hypotheses rather than normative spec text.

The note ends with five specific falsification targets that should be checked before freezing an SPN formalism.

MIT / DCO sign-off included in the commit.