[TS PBT] Add core development corpus and pilot protocol

[CaelmBleidd]

Problem

#405 needs a development corpus and preregistered comparisons before the core feedback loop is measured. Without pinned source oracles, capability boundaries, and negative controls, pilot outcomes can be selected after seeing results.

This checkpoint

• Add development v1 with two independent external property families: es-toolkit chunk order/shape and fast-check duplicate removal. Pin source revisions, MIT notices, original oracle meaning, native versus bounded generators, and expected capability limits.
• Include a correct implementation for each external family, the upstream intentionally faulty duplicate-removal example (a controlled fault, not a real library defect), and curated parity, passing hypothesis-refutation, and no-benefit mechanisms.
• Validate correct original-runtime behavior and explicit witnesses. Classify floating-point round-trip equality as a false specification.
• Record development/held-out split rules, adaptation estimates, a six-configuration equal-budget pilot protocol with planned controls, and a provisional nearest-work/claim matrix.

Verification

At bf397b27e3197bc09ba6e461d21e80b909afb566:

• npm ci --ignore-scripts --no-audit --no-fund in usvm-ts-pbt/corpus/development/v1: passed.
• npm --prefix usvm-ts-pbt/corpus/development/v1 run validate: passed (500 deterministic generated inputs per correct property plus explicit witnesses).
• git diff --cached --check: passed before commit.
• JSON manifest case-ID consistency: 6 IDs matched 6 cases.

Limits and next work

This is an early input for #405, not completion of #356, #357, or #404. No comparative search, exact branch saturation check, symbolic capability admission, measured adaptation time, held-out selection/results, real known defect, or #400–#403 strata are claimed. #395 owns executable PropertyManifest registration and may mark cases concrete-only; unsupported modes must remain outside the paired denominator. The curated parity fixture is only a saturation candidate until exact runtime branch evidence confirms it. Literature novelty is provisional; no manuscript or public thesis file is included.

Refs #345 #356 #357 #404 #405