fix: wrap explain_query in the Raw { engine, format, payload, ... } shape (#89)

[aesslinger]

Summary

• explain_query returned the raw EXPLAIN JSON value directly, with a comment acknowledging the host would fall through to ExplainQueryOutput::Plan { plan: res } instead of the Raw variant the builtin driver emits.
• The host's plugin adapter (tabularis plugins/driver.rs::explain_query) only classifies a response as Raw when it finds engine/format/payload as strings (via .as_str()) plus an optional original_query — any other shape, including a bare JSON array, falls through to Plan. Runtime-registered EXPLAIN parsers select on engine+format, so this plugin's output was never selectable by them, and the parsed-plan renderer ran instead of the raw-payload path the builtin uses.
• Wrapped the plan JSON in the exact shape the adapter checks: { engine: "postgres", format: "postgres-json", payload: <stringified JSON>, original_query: <the query> }. payload must be the JSON string form, not the live JSON value — the adapter reads it with object.get("payload")?.as_str(), which silently returns None (not an error) for a nested object/array, dropping the response into the Plan fallback exactly like the pre-fix bare-array response did.

Fixes #89.

Test plan

• TDD: added raw_explain_output unit tests to query_tests.rs, including one that specifically checks payload is Value::String (not a nested value) to catch that exact silent-fallback failure mode. Confirmed the new tests fail to even compile against the pre-fix code (the function didn't exist yet) before the fix.
• cargo test --lib — 327 passed
• cargo clippy --all-targets -- -D warnings — clean
• cargo fmt --all -- --check — clean
• Verified live against a real PostgreSQL instance by simulating the host adapter's exact matching logic (object.get(field)?.as_str() for each of engine/format/payload) against real explain_query responses:
  • Confirmed the pre-fix binary's bare-array response is not classified as Raw (falls through to Plan, reproducing the bug)
  • Confirmed the post-fix binary's response is classified as Raw, with engine/format/original_query correct and payload containing a real EXPLAIN plan
  • Confirmed this holds both without and with analyze=true, and that real execution stats ("Actual Rows") come through correctly in the ANALYZE case

Note for tabularis (not part of this PR)

While reviewing the cross-repo parity suite (src-tauri/tests/postgres_integration/parity_explain.rs), noticed its top-of-file comment says: "ExplainQueryOutput differs structurally between built-in (Raw variant) and plugin (Plan variant)" — documenting the exact divergence this PR fixes. The tests themselves only assert Ok (not the response shape), so they'll keep passing either way, but that comment is now stale once this merges. Flagging for awareness; not something to fix in this repo.

[github-actions]

Version suggestion

Based on this PR's title (fix) and the prerelease:rc label:

Current	1.0.0-rc.3
Suggested next tag	v1.0.0-rc.4

This is informational only — no tag or release is created automatically yet.