test: recursion_limit=1024 on remaining 22 test files (final fan-out)

[AdaWorldAPI]

Summary

Final fan-out of the #[recursion_limit = "1024"] pattern started in Sprint O (#1, tests/create.rs) and Sprint S (#5, 7 CRUD files). Brings all 30 surrealdb/core/tests/*.rs files to a uniform state so the full upstream integration suite can be exercised under SURREAL_TEST_KV=lance without compile-time recursion errors.

Files (22)

access.rs       cache.rs        function.rs    param.rs    sequence.rs
alter.rs        changefeeds.rs  future.rs      query.rs    table.rs
asyncevent.rs   complex.rs      index.rs       remove.rs   timeout.rs
auth_limit.rs   define.rs       info.rs        script.rs   util.rs
field.rs        live.rs

Why

When both kv-lance and kv-mem features are enabled (the combination is needed for SURREAL_TEST_KV=lance routing via helpers.rs::new_ds, because upstream iam/file.rs requires tempfile which is kv-mem-gated), the async-trait expansion in surrealdb_core::expr::expression::Expr::compute exceeds rustc's default recursion limit of 128.

Same compile-time error pattern as #1 + #5:

error: queries overflow the depth limit!
  help: consider increasing the recursion limit
  note: query depth increased by 130 when computing layout of
        {async fn body of surrealdb_core::expr::expression::Expr::compute()}

Mechanical change

Each file gets a single line: #[recursion_limit = "1024"] prepended to line 1. Rust permits inner attributes in any order, so files that already had #[allow(...)] or #[cfg(...)] inner attrs are unaffected. Zero runtime cost.

Coverage after this PR

Sprint	PR	Files
O	#1	create.rs
S	#5	select.rs, update.rs, delete.rs, insert.rs, upsert.rs, merge.rs, relate.rs
V	this	22 files (the rest)

Total: 30/30 integration test files.

Test plan

• CI default build untouched (no feature change, prepend is pure inner-attribute).
• cargo check --tests --features "kv-lance kv-mem" --no-default-features compiles all 30 test binaries (verified locally; running in background at PR-creation time).
• Behavioral pass-rate signal per test file under SURREAL_TEST_KV=lance is the next-sprint deliverable; that data will be aggregated into .claude/lance-backend/KNOWN_DIFFERENCES.md as semantic gaps are found.

Followups

After this PR, the open KNOWN_DIFFERENCES items are:

• Mixed writes+deletes commit non-atomic (Lance API limitation; sequential).
• Multi-bucket BindSpace sharding (Phase 2 throughput; architectural).
• lance-graph Cypher engine as SurrealQL function (Phase 3).
• blasgraph GraphBLAS algebra (Phase 3).
• Benchmarks vs RocksDB / SurrealKv.
• Per-file pass rate aggregation under SURREAL_TEST_KV=lance.

---

Generated by Claude Code

[AdaWorldAPI]

Meta-V verified locally

cargo check --tests --features "kv-lance kv-mem" --no-default-features
→ Finished `dev` profile in 2m 59s, 0 errors

All 30 test binaries (including the 22 in this PR) compile cleanly under the combined kv-lance + kv-mem feature set. The remaining warnings (e.g., (test "remove") generated 1 warning) are pre-existing and unrelated to the recursion_limit additions.

Ready for merge.

---

Generated by Claude Code