release: scheduling-kit 0.10.0 — availability substrate (TIN-1996)

[Jesssullivan]

Release: scheduling-kit 0.10.0 — availability substrate (TIN-1996)

Cuts the release that makes the availability substrate consumable off the
tinyland-inc/bazel-registry SSOT and the derived GitHub Packages artifact.

Why 0.10.0 (MINOR)

Since the last release (0.9.2), the following landed on main:

• feat(adapters): availability substrate — RRULE windows minus CalDAV busy overlay (TIN-1996) #114 — availability substrate: two net-new public subpath exports
  (./caldav → caldav-busy, ./availability → availability-substrate)
  plus new inter-module deps (tummycrypt_tinyland_calendar,
  tummycrypt_tinyland_caldav_client) mirrored as exact-version @tummycrypt
  npm devDependencies (TIN-1996).
• feat(kit): availability snapshot freshness attribution (TIN-945 row 3) #116 — availability snapshot freshness attribution (TIN-945 row 3):
  net-new additive public API on the existing ./availability (./adapters)
  subpath — getAvailabilitySnapshot plus AvailabilitySnapshot /
  AvailabilityFreshness / SnapshotFreshnessPolicy and
  isSnapshotStale / isSnapshotExpired / isSnapshotReusable. Thin wrapper
  over getSubstrateAvailability (feat(adapters): availability substrate — RRULE windows minus CalDAV busy overlay (TIN-1996) #114) that stamps observedAt / staleAt /
  expiresAt + cold-read readMs so a cache consumer can attribute cold-read
  cost and reason about reuse. No new subpath, no change to existing exports —
  additive-only, reinforcing the MINOR bump.
• test(integration): cassette-backed idempotency replay proof (TIN-1995) #115 — cassette-backed idempotency replay proof (TIN-1995).
• ci(bazel): enroll scheduling-kit in the shared cache-backed lane (TIN-2110) #111–docs: SchedulingAdapter parity matrix — 17 methods, real vs advisory with citing test IDs (prompt-03) #113 — Bazel cache-lane enrollment, ci-templates v2.5.0 gate, and the
  SchedulingAdapter parity matrix.

Net-additive public API surface with no breaking changes to existing exports is
a SemVer MINOR bump: 0.9.2 → 0.10.0. The repo has no CHANGELOG.md and no
written SemVer policy that overrides this; release truth lives in the version
triple + AGENTS.md + the generated docs, all updated here.

Rebased onto main after #116 landed. This branch was rebased on top of
the merged #116 commit and generate-doc-artifacts was re-run, so the 0.10.0
generated package-surface.md reflects the source actually on main
(src/adapters file count 14 → 15, picking up the new snapshot suite). The
release description above and the validation below match the landed surface.

Version triple + surfaces (aligned)

Surface	Value
package.json version	0.10.0
MODULE.bazel module version	0.10.0
BUILD.bazel //:pkg npm_package version	0.10.0
AGENTS.md current-released / registry truth	0.10.0 / 0.10.0+
README.md bazel_dep + registry truth	0.10.0 / 0.10.0+
docs/generated/{package-surface,release-metadata}.md	regenerated → 0.10.0

Validation (all green, run against this rebased branch)

• pnpm check:release-metadata → release metadata aligned for @tummycrypt/scheduling-kit@0.10.0 (includes the in-house Bzlmod↔npm parity loop)
• node scripts/generate-doc-artifacts.mjs --check → generated docs in sync
• pnpm test:unit → 860 passed (33 files) — includes feat(kit): availability snapshot freshness attribution (TIN-945 row 3) #116's 10 snapshot acceptance tests
• pnpm test:integration → 34 passed (3 files)
• pnpm check (svelte-check) → 0 errors, 0 warnings
• pnpm build → src -> dist OK; ./caldav and ./availability resolve to real dist/adapters/*.js + .d.ts
• pnpm check:package (publint) → All good!

Publish handoff

Per the delivery doctrine, the git tag is the publish trigger and stays an
operator step. Do not push a tag from this PR.

merge then operator tags v0.10.0 to publish.

npmjs stays retired/frozen at 0.8.0 (npm_publish_mode: disabled); delivery is
the Bzlmod module graph (SSOT) + derived GitHub Packages @jesssullivan/scheduling-kit.

Closes the release-prep slice of TIN-1996.