docs: rewrite project overview and surface underexposed research

[LightChainr]

Purpose

Rewrite the repository entry point around the scientific structure that is actually supported by the full research history, and make substantive but easy-to-miss work visible without confusing integration state with evidence maturity.

The old README was anchored to an August 29 execution snapshot: it still described N145→290 as future work and over-centered the H4/Q4/Jordan narrative. This PR replaces that framing with a durable project overview.

README rewrite

README.md is now organized around:

• the exact ambient-homology source X=r-1, with canonical A_top=<X> and E_top=<X^2> coordinates and the exact X^3=X source algebra;
• the inference stack topology/semantics -> sufficient statistics -> prospective evidence -> mechanism elimination -> identifiability -> continuum naming;
• a compact established-vs-unresolved table separating strong finite-size evidence from field/operator identification;
• the prospective N185/N265 and norm-5 evidence, plus the completed N145→290 full-curve failure of one-scalar closure;
• project-changing negative results: scalar correction failures, generator-dependent state dimension, branching nonclosure, P398 reversible controls, channel/covariance QA;
• the current live frontier: [P0 theory] Original Matching-One observable: homology-marked trace, normalization, and identifiable predictions #275 P0 original-observable identifiability and [P1 theory] Full original-U thermal/contact asymptotics after the completed one-defect gate #537 P1 proof-level thermal/contact obligations;
• first-class sections for P334 predictive-state geometry, P398 hidden finite transfer, primitive C3, rigorous threshold work, and computational statistics;
• collapsible summaries of the bounded degree-1..4 threshold-relation census and the mature [P2] Automated self-dual gadget and critical-manifold search #13/[P2] Search replacement gadgets for improved rigorous bounds #14 finite terminal algebra;
• explicit reading rules, evidence discipline and nonclaims.

Research visibility atlas

Adds docs/RESEARCH-ATLAS.md, a deeper map of underexposed work:

• branch-only/open-PR/closed-unmerged science;
• negative results and semantic corrections;
• P334 continuation/cut-network and P398 hidden-geometry work;
• P250 generator-dependent realization results;
• P321/P215/P40/P337/P439 result chains;
• original-U statistical reachability and thermal/pivotal proof gates;
• Issue [P2][NUM] Bounded PSLQ / integer-relation exclusions after threshold provenance reconciliation #1 bounded relation-search census;
• Issue [P2] Automated self-dual gadget and critical-manifold search #13/[P2] Search replacement gadgets for improved rigorous bounds #14 mature exact algebra and theorem-facing boundary;
• explicit nonclaims and dependency/evidence-role distinctions.

docs/RESEARCH-MAP.md gains a visibility rule directing overview work to the atlas before compressing the project into the compact track map.

Boundary

This PR changes documentation/navigation only. It does not:

• modify raw data, frozen predictions, scorers, result artifacts or claim levels;
• merge or promote branch-only/open-PR results;
• change Issue priorities or authorize production;
• turn the README/atlas into a second execution queue.

The intended distinction is explicit throughout: integration state, evidence role, model compatibility and physical identification are different coordinates.

Diff scope

Three files only relative to main: README.md, docs/RESEARCH-MAP.md, and new docs/RESEARCH-ATLAS.md.