docs: honest framing — tamper-evident/verifiable, not 'deterministic/reversible like Git'

[New1Direction]

Realigns the project's headline positioning to what it actually is — and to the live site's own framing (tamper-evident, verifiable, zero trust, no blockchain).

Why

The published korg crate description and the README / ROADMAP / CONTRIBUTING taglines all claimed "the first deterministic cognitive runtime … reversible — like Git, but for cognition." But:

• the codebase stripped "revert theater" in f04ccfe ("honest dashboard — strip fabricated telemetry, git history & revert theater"), and
• LLM agents are not deterministic.

So the headline overclaimed against the project's own honest direction.

What changed (surgical — tagline only)

Spot	Before → After
Cargo.toml description	"first deterministic cognitive runtime … reversible like Git" → "a tamper-evident, verifiable ledger for AI agent cognition … anyone can independently verify — no blockchain, no trust in the tool that produced it"
Cargo.toml keywords	deterministic → verifiable, tamper-evident
README.md subtitle	"reversible — like Git, but for cognition" → "recorded in a hash-chained ledger you can independently verify — tamper-evident, zero trust"
ROADMAP.md / CONTRIBUTING.md	same tagline realigned

Deliberately left untouched: the accurate technical uses of "deterministic" — JCS canonicalization, projection folds, journal replay, git-read-tree rollback, HLC ordering, the embedding test fixture. Those genuinely are deterministic; only the cognition-level headline was the overclaim.

Also normalizes the @korgg/ledger-verify package.json repository.url (silences the npm publish warning).

Docs/metadata only; cargo verify-project OK. Left unmerged for your review — it's brand copy, your call. (crates.io's korg@0.1.0 page only updates on the next republish.)