Exact source snapshots for the FIRST frozen purchase latency run.

The primary outcome file is results/purchase_postgresql_initial_lookup.json.
It includes all 540 raw timing samples, 22 correctness checks, parameters, query
plans and SHA-256 hashes of the executed source files. This first implementation
used a JSONB set-returning-function join for its three epoch lookups. Its measured
latency was worse than the matched native control and is retained as a negative
result, irrespective of later performance work.

These files are archival snapshots, not an independently importable package.
To rerun the initial implementation, make an isolated copy of the full v0.4
project, replace that copy's v04/purchase.py with this purchase.py and replace
its experiments_v04/run_purchase.py with this run_purchase.py. Verify the hashes
against the initial result before running on a fresh dedicated PostgreSQL
installation, using the same documented parameters. Do not modify the primary
project in place or target an unknown database.

Any optimization informed by this run is development evidence, not an
independent performance holdout or a preregistered novelty confirmation.
