docs(research): record compact u4 bit-table negative result - #85

Review recommendation: close or replace this version.

The negative result claims successful correctness experiments and optimizer timeouts without retaining the experiment implementation or sufficient reproduction commands/configuration. Close this version or supply the code, deterministic inputs, compiler policy, timing configuration and explicit failure criteria so the result is evidence-backed and reproducible.

September 11 local validation (unchanged PR head): python3 tools/kb.py validate passed.

Reviewed commit: d0eaf382201197b63def783707a6c9a443128f52.

Closing as requested by review. This negative result does not retain the experiment implementation, deterministic fixtures, exact compiler/optimizer policy, timing configuration, timeout criteria, or logs needed to reproduce the claims. The reported 235-nibble capacity is only the arithmetic bound (4×235 + 60 = 1000), not a measured peak including branch temporaries, and the claimed optimizer slowdown is not separated from interpreter execution. The proposed NR-043 identifier is also already occupied on current main. Reopen only with recovered evidence and a reproducible rerun.