docs: agent-facing documentation for sei-load

[bdchatham]

Comprehensive agent-facing documentation so a cluster of performance-benchmarking + Sei-expert agents can autonomously configure, run, reproduce, and interpret experiments with sei-load.

What

• Root AGENTS.md index (entry point + reading order + fastest-path).
• docs/ tree (8 sections): mental model (open-loop / coordinated omission; generator emits signals, verdicts are external), running, config reference, workload model (+ what the contention/size axes actually probe on Sei — block-stm/OCC, not DynamoDB-style throttling), reproducibility (seeded sub-streams, the honest determinism guarantee), measurement & metrics (exact emitted-series catalog + external-query guidance), experiment playbook (objective → knobs → run → interpret), limits & accepted boundaries.
• README banner pointing agents to AGENTS.md.

How it was built

Authored by matched experts grounded in source, then a full Coral cohort cross-review: systems-engineer (accuracy), sei-network-specialist (Sei mechanism), opentelemetry-expert (metric catalog + PromQL), prose-steward (agent-usability). Notable catches folded in: the PromQL examples now use the real seiload_-prefixed wire names (the catalog was right but queries were querying base names → zero series), and a false seiChainId 'casing bug' claim was removed (Go binds case-insensitively — verified).

⚠️ Sequencing

§04/§07 document the StorageRW key/size/op contention axes, which are delivered by PLT-465 (#54, currently open) — on main those config fields parse but don't affect generated txs. The docs mark this clearly; this PR should merge with or after #54 so the documented axes exist on main.

Docs-only; no code changes.

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Markdown-only additions with no changes to load generation, sending, or metrics emission; main risk is doc drift or misleading guidance if behavior changes without doc updates.

Overview
This PR is documentation only — no runtime or config code changes.

It introduces AGENTS.md as the entry point: reading order, a table of contents, a fastest-path to a first experiment, and standing caveats (PLT-465 StorageRW gaps, missing schedule_lag metric, text-only --report-path, config-only seed, seiload_ Prometheus wire names, external verdicts).

The docs/ set covers the full agent workflow: 01 send pipeline, open- vs closed-loop, coordinated omission, conservation identities, measure-not-judge; 02 build, CLI flags, lifecycle, sample invocations; 03 JSON LoadConfig / settings / scenarios / funding and schema-vs-main gaps; 04 scenario catalog and StorageRW axes (flagged as PLT-465); 05 seed → frozen RNG sub-streams and fair A/B; 06 conservation model, emitted metric catalog with seiload_-prefixed PromQL, and external validity heuristics; 07 experiment playbook (objective → knobs → validity → interpret); 08 accepted measurement boundaries and which counters discount results.

README gets a banner directing agents to AGENTS.md and noting the README may lag.

Docs repeatedly warn that on main, StorageRW distribution/size/op config parses but does not drive txs until PLT-465 (#54) lands, and that agents must compute rates/percentiles/SLOs themselves from emitted signals.

^{Reviewed by Cursor Bugbot for commit b6f9b15. Bugbot is set up for automated code reviews on this repo. Configure here.}