Execute SQL against a Turso Cloud database over raw HTTP (/v2/pipeline) without using an SDK

Documented steps

1. Get the database base URL: https://[databaseName]-[organizationSlug].turso.io (from `turso db show <db> --http-url`).
2. Mint a data-plane auth token: `turso db tokens create <db>` (add `-r/--read-only` for read-only).
3. POST to <base-url>/v2/pipeline with headers `Authorization: Bearer <authToken>` and `Content-Type: application/json` (https://docs.turso.tech/sdk/http/quickstart).
4. Body is a batch of typed requests: {"requests":[{"type":"execute","stmt":{"sql":"SELECT * FROM users"}},{"type":"close"}]}.
5. Bind parameters instead of interpolating: "args":[{"type":"integer","value":"1"}] for positional, or named_args for named. Valid arg types are null, integer, float, text, blob.
6. Encode every scalar value as a JSON STRING ("value":"1", not 1) — the protocol does this deliberately to avoid float/int64 precision loss.
7. Read results from the response's results array; each execute returns cols, rows and affected_row_count.
8. Include {"type":"close"} as the last request for one-shot calls so the server tears down the stream; omit it and keep the returned `baton` if you intend to issue follow-up requests on the same stream.
9. For anything beyond simple calls, prefer an SDK (@tursodatabase/serverless, libsql) rather than hand-rolling the protocol.