chore: prepare v0.2.2 manifest alignment and release notes (COD-476)

[shivros]

COD-476 — Hydra v0.2.2 manifest alignment and release notes (preparation only)

Re-prepared on endver after the hosted runner's worktree (and its commits) were lost to /tmp cleanup on builder-01. Same spec, fresh implementation, commit 232e168.

What changed

• Cargo.toml: workspace.package.version 0.1.0 → 0.2.2 (all five members inherit)
• Cargo.lock: exactly 5 workspace package version lines updated — zero registry/source/checksum changes
• README.md: new Releases & versioning section — v0.2.2 marked proposed, not published; immutable historical-tag policy (v0.1.0–v0.2.1 keep their 0.1.0 manifests); honest feat: add DeepSec security-scan tracer bullet (COD-450) #6–test: cover unary cursor pagination projections #8 summary (reference/metadata coverage, no production changes under crates/*/src); git-tag consumer pin examples with regen requirement, explicitly valid only after publication ("until then pin v0.2.1")

Not in this change

No tag, no GitHub release, no crates.io publication, no consumer repins. Publication remains a separate task requiring exact-version release authorization (tags are immutable).

Verification (all on endver, base a68c3e5)

• cargo metadata --locked --no-deps: all 5 workspace members report 0.2.2 (hydra-core, hydra-codegen, hydra-mcp-stdio, notes-example, security-scan-example)
• Lock diff vs baseline: 5 insertions / 5 deletions, version lines only
• cargo build --all-targets ✅ · cargo test --all-targets ✅ 58 passed, 0 failed
• cargo clippy --all-targets -- -D warnings ✅ · cargo fmt --all -- --check ✅
• Both examples: check → snapshot → write ×2 → byte-compare → check — all six artifacts byte-identical across both passes
• creed diff ✅ · git diff --check ✅ · diff limited to Cargo.toml, Cargo.lock, README.md

Closes COD-476

[shivros]

🤖 Automated Review Panel

Dual-model review before opening. Both seats APPROVE; no blockers, no minors outstanding.

Reviewer A

Verdict: APPROVE — 3 nits, none blocking:

• [nit] Consumer pin example shows tag = "v0.2.2" inline; a skimming reader could copy it before the tag exists — mitigated by the bolded status line and the explicit "until then pin v0.2.1" sentence.
• [nit] Regen instruction assumes the consumer's default definition path; consistent with the repo's own documented usage.
• [nit] hydra-codegen pin example given by reference ("same shape") rather than a literal second snippet.

Reviewer B (independently verified claims)

Verdict: APPROVE:

• Cargo.toml: workspace version = "0.2.2"; all 5 members confirmed on version.workspace = true.
• Cargo.lock diff is exactly the 5 workspace version pairs — no source/checksum/dependency changes; cargo metadata --locked --offline passes.
• README truthfulness independently confirmed: no v0.2.2 tag exists locally or on GitHub (tags stop at v0.2.1, releases list empty); hydra-codegen 404s on crates.io; PR feat: add DeepSec security-scan tracer bullet (COD-450) #6/chore: adopt layered agent context #7/test: cover unary cursor pagination projections #8 subjects match the README descriptions and none touched crates/*/src; historical tags v0.1.0–v0.2.1 all had manifest 0.1.0.
• Generated artifacts unaffected: check passes in both examples after the bump.
• [minor→verified-not-a-defect] All "proposed, not published" claims verified truthful.

No tag, release, or publication is created by this PR. Publication remains a separate authorization-gated task.

[shivros]

Auto-Merge Gate

Confidence: 0.94

This narrowly aligns Hydra's inherited workspace manifest version and lock records to the proposed v0.2.2 release, and documents the status accurately. The diff is limited to Cargo.toml, the five corresponding workspace entries in Cargo.lock, and release/versioning documentation. It creates no tag, GitHub release, crates.io publication, consumer repin, deployment, or cutover.

Observed checks:

• Test — SUCCESS
• Creed context drift — SUCCESS

Linked Linear ticket COD-476 was checked for manual-review / do-not-merge markers; none were found. The review panel approved both independent reviews with no blockers.