0.10: Produce release evidence and frozen-candidate closeout

[bazer]

Objective

Own the complete DataLinq 0.10 evidence lifecycle: freeze the 0.9 before-state before shared runtime work, add attributable workstream gates as implementation lands, and produce one reproducible frozen-candidate go/no-go handoff without publishing anything.

Required scope

• Capture the W0 repository, SDK, package, provider, test-catalog, API, generated-source, benchmark, telemetry, and I/O-boundary baseline.
• Record command lines, exit codes, timestamps, environment identity, artifacts, warnings, and dispositions.
• Add focused evidence for A10, H10, V10, T10, and G10 as those contracts land.
• Run the complete supported provider/test matrix and validate its expected inventory.
• Run ApiCompat, generated-code comparison, package inspection, external consumer smoke, and applicable AOT/trim/WebAssembly/browser checks.
• Compare performance and telemetry to a same-runner 0.9 baseline without hiding issue Restore DataLinq 0.9 allocation efficiency to the 0.8 baseline #26 or attributing new regressions to old debt.
• Validate public documentation and internal links against shipped behavior only.
• Freeze one exact candidate, pack without publishing, write the release manifest, and record an explicit GO or NO-GO.

Acceptance criteria

• The W0 baseline is complete before A10 or G10 changes shared execution or generator behavior.
• Every required command has a tracked successful exit and complete final summary, or an explicit blocking failure; partial evidence is not reported as verified.
• Every expected suite, provider target, package, compatibility lane, and documentation gate is present and attributable to one candidate commit/version.
• No undispositioned API break, package finding, compatibility failure, warning, or material performance regression remains.
• Public documentation describes only behavior proven by the frozen candidate.
• manifest.md identifies exact artifact hashes and records an explicit GO or NO-GO.
• Tagging, package publication, GitHub release creation, and announcement remain separate maintainer actions.

Dependency and order

This issue starts first with RE10-0/W0, remains open throughout implementation, and closes last after every other required 0.10 workstream is complete:

• 0.10: Implement native async and cancellation #107 — A10 native async and cancellation
• 0.10: Add DI, hosting, and explicit unit-of-work support #108 — H10 DI, hosting, and explicit unit of work
• 0.10: Add opt-in startup schema validation #109 — V10 startup schema validation
• 0.10: Add application testing builders, fixtures, and unit-of-work fakes #110 — T10 application testing support
• Support source type aliases in generated models #93 — G10 source type alias correctness
• W8 provisional integration evidence
• W9 frozen-candidate closeout and explicit GO/NO-GO

Explicit non-goals

• Publishing packages, tagging a release, or creating a GitHub release.
• Treating final-0.8 allocation parity from Restore DataLinq 0.9 allocation efficiency to the 0.8 baseline #26 as a hidden 0.10 release gate.
• Adding any excluded roadmap item because implementation finishes early.

Planning sources

• 0.10 implementation roadmap
• Implementation order and integration plan
• Release evidence and closeout plan