P1: Release: pick the version and publish the three crates to crates.io

[macanderson]

Problem

The README and docs link crates.io/crates/contextgraph-types (et al.) as if live, and PUBLISHING.md contains the full release runbook — but the crates do not appear to be published yet (cargo search contextgraph-types returns nothing). Until cargo add contextgraph-types works, every "build a provider today" claim in the README dead-ends at a 404, and the freeze criterion of independent implementations has nothing to build against.

Proposal

Execute the PUBLISHING.md runbook for 0.1.0:

• Pre-flight: cargo publish --dry-run for each crate in dependency order (contextgraph-types → contextgraph-host → contextgraph-conformance); verify each crate's publish = true opt-in, license files, description/keywords/categories metadata, and README rendering on the crates.io page
• Verify docs.rs builds (features, no doc warnings under --cfg docsrs if applicable)
• Move CHANGELOG.md [Unreleased] → ## [0.1.0] - <date>
• Tag v0.1.0, publish in dependency order
• Post-flight: cargo add contextgraph-types in a scratch project compiles; README badges (crates.io version, docs.rs) added
• Decide the release cadence for pre-freeze 0.x bumps (per docs/stability.md, any 0.x → 0.y may break — batch normative changes into planned minors rather than trickling)

If the crates are in fact already published under names I could not see, close this after verifying the links and adding the badges.

Sequencing note: worth landing CI first (see the CI issue) so the release tag is cut from a commit the suite has verified, and ideally publishing via a tag-triggered workflow with a crates.io trusted publisher/token rather than a laptop.

[macanderson]

Re-verified 2026-07-26 against origin/main @ 1ee9b1c.

All four workspace crates 404 on crates.io and on the sparse index. There are zero releases, and the only tag in the repo is the pre-rename ocp-v0.1.0. The README still links three live-looking crates.io URLs, every one of which 404s — so "build a provider today" does not merely dead-end, it points at a URL that looks published and is not.

Recommended scope: keep, with two edits — the CHANGELOG has already consumed [0.1.0], so this now needs an explicit version decision before anything is published; and drop the "land CI first" note, since CI now exists (#2's residue is the merge gate, not the pipeline).