Update Shilin H4 and agentic release artifacts

[Owen-1234]

Summary

• Add bounded H4 early-wallet buyer/holder proxy backfill support and publish the current 10-token decoded-proxy sample.
• Add registered multi-model agentic scaffold ablations, including no-API local Ollama runs for Llama 3.1, DeepSeek-R1 8B, and Qwen2.5 0.5B.
• Refresh Shilin release artifacts, README, Dataset Card, Schema Contract, data dictionary, requirement closure audit, top-conference gap ledger, and tests.
• Keep claim boundaries explicit: H4 remains a proxy sample, Telegram Case B remains matched/predictive rather than causal, and Base remains bounded until full archive/indexer coverage is complete.

Key Results

• H4 early-wallet sample: 10 tokens, 298 parsed early pool transactions, 101 classified buyer/seller/holder proxy transactions, 53 buyer-proxy wallets, and 60 holder-proxy wallets.
• Agentic ablations: 65 scored cells and 105 successful outputs across DeepSeek deepseek-chat, Ollama llama3.1, Ollama deepseek-r1:8b, and Ollama qwen2.5:0.5b.
• Top-conference gap ledger now reports partial_same_cohort_rpc_buyer_holder_proxy_10_tokens_298_parsed_txs_101_classified_txs and partial_4_model_specs_65_scored_cells_105_ok_runs.

Validation

• python3 scripts/build_benchmark_release.py
• python3 scripts/check_artifacts.py
• python3 -m unittest discover -s tests -v

All 25 tests passed locally.