[VikingMew]

Summary

• add a reproducible one-off process-level benchmark harness for all six QPY-83 workloads
• commit the exact jiter fixture, shared workload sources/templates, raw samples, environment metadata, and derived comparison table
• report five checksum-verified comparable workloads and classify json_dumps as a semantic-difference failure without inferring performance

Test Plan

• Run python3 benchmarks/talk_comparison/run.py with CPython 3.11.16 after a single cargo build --release
• Verify the committed fixture SHA-256 matches the upstream jiter fixture
• Recompute sample counts and derived min/average/per-iteration values from raw-results.json
• Run python3 -m py_compile for the harness and committed Python workloads
• Run cargo fmt --all -- --check
• Run cargo clippy --workspace -- -D warnings
• Run cargo test --workspace (585 passed, 59 ignored)

Fixes QPY-83