README: promote the honest scope-coverage table (+ interface summary) to a user-facing artifact

[paul-hammant]

Motivation

blazerules opens its README with a 4-row Interface summary table (Python SDK / CLI / agent / dashboard — "what it is / entry points") and carries a scope-coverage mindset throughout its user-facing docs. It earns trust by stating capability boundaries plainly to users.

aeb already has the best version of this table anywhere — the scope-scoring table in LLM.md with ✓/◐/✗ marks and unembellished per-dimension status. The problem: that honest self-assessment lives in a notes-to-self file a user never reads, while the README undersells what's actually done and hides what's still ✗ TODO.

Proposal

Promote the scope-coverage table (or a user-facing derivative) into README.md as a first-class section.

• Keep the three-tier honesty: ✓ Done / ◐ Partial / ✗ TODO.
• Trim the LLM-facing implementation asides; keep "what good looks like" + "aeb status."
• Add a short Interface summary table above it (à la blazerules): the ways to invoke aeb — aeb <target>, aeb --scan, aeb --graph, aeb --since, aeb gcheckout, aeb --sandbox/--vet, aeb --watch — one row each, "what it is / entry point."
• LLM.md keeps its own copy with the implementation detail; the README version is the trimmed public one. A note in each points at the other so they don't silently drift.

Acceptance criteria

• README has an Interface summary table listing every invocation surface with a one-line description.
• README has a scope-coverage table with ✓/◐/✗ status per dimension, derived from the LLM.md table.
• Both tables are honest — ✗ TODO rows (remote cache, lockfiles, publishing, IDE/LSP, cross-compile, CI integration) are shown, not omitted.
• Cross-reference note in both README.md and LLM.md so the two copies are known-linked.

Not being asked

Not asking to auto-generate the table from code, or to make the two copies a single source with an include mechanism. Hand-maintained + cross-referenced is fine at this size.