research: analyze July Zenith mission latency

[apodgaiko]

What changed

Adds a reproducible postmortem of the July 17-19 production-log mission, including:

• a five-minute schematic explanation of the Zenith harness;
• a timestamped execution ledger and quantitative cost profile;
• causal analysis of task growth, validation repetition, evidence amplification, and terminal-review failure;
• a minimum-sufficient counterfactual mission DAG;
• a ranked, branch-ready harness improvement backlog.

Why

The mission took 50.54 wall hours and grew from an inferred 24 tasks to 168. This research separates user-requested scope from harness-created overhead and proposes measurable changes that preserve correctness, traceability, independence, and explicit limitations.

Validation

• Ruff check passed
• Python compilation passed
• Generated metric invariants passed
• Repeated analysis produced deterministic outputs
• Git whitespace checks passed

This PR is research and technical proposals only; it does not yet change runtime behavior.