ci: use Spicepod v2 in the test fixture

[lukekim]

What

Changes version: v1beta1 to version: v2 in test/scripts/spicepod.yaml. One line.

Why

Current spiced rejects v1beta1 outright:

Failed to start Spice runtime: Unable to load spicepod .../test/scripts:
Unsupported Spicepod version in .../test/scripts: 'v1beta1'

Supported versions are: v1, v2 (or full semver: v1.0.0, v2.0.0-rc.1, etc.)

With no spicepod loaded the runtime never starts serving, so every Local Runtime test fails with connect ECONNREFUSED 127.0.0.1:8090 — 15 failed / 3 passed per matrix entry.

This is repo-wide, not specific to any change: PRs #311, #312 and #313 all fail identically. It's version drift between the fixture and the installed runtime, which will keep every PR red until it's fixed.

Verification

Ran spiced locally against the amended fixture (verified under both v1 and v2; v2 is what spice init emits, so the fixture now matches the CLI rather than trailing it). The Unsupported Spicepod version error is gone, both datasets are picked up, and the runtime serves:

runtime::init::dataset: Loading datasets: 2 tasks dispatched, 0 skipped (of 2 total)
runtime::init::dataset: Dataset test_postgresql_table_not_accelerated initializing...
runtime::init::dataset: Dataset test_postgresql_table_accelerated initializing...
runtime::http: Spice Runtime HTTP listening on 127.0.0.1:8090

Both datasets then fail to connect to PostgreSQL, which is expected — there's no postgres on this machine. CI provides one via its services: block, so that part is exercised there.

Deliberately scoped to the one line; no other fixture or workflow changes.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
spice-js	Ready	Preview	Jul 27, 2026 10:18pm

[claudespice]

The six red Cloud Tests legs on run 30310185390 are not this diff — they are a transient outage of the deployed endpoint the test hits, and a rerun should clear them.

Evidence:

• Every failure is the same one: HTTP query failed with status 500: {"success":false,"error":"undefined undefined: undefined"} from cloud › Vercel Endpoint (via SDK). The five failing cases are all plain queries (constants, data types, string ops, math ops, large result sets); error handling works passes in both suites, because it only asserts that a bad query errors — which a blanket 500 also satisfies. That shape is a dead endpoint, not a behavior change.
• This PR changes only the test fixture's Spicepod version. Nothing in it reaches the deployed endpoint.
• The same six jobs went green on 2026-07-30 in run 30584139765, three days after this run, with no change to them.

I can't rerun it myself (Must have admin rights to Repository), so this needs a maintainer's click: Re-run failed jobs on run 30310185390.

Worth clearing — #314 is waiting on this one to merge, and its own CI is separately parked at action_required.