fix: parenthesize IN-list quals when deparsing

[breken-ai]

What kind of change does this PR introduce?

Bug fix (qual deparse, supabase-wrappers/src/interface.rs and mysql_fdw).

What is the current behavior?

Qual::deparse_with_fmt renders an IN list (use_or = true) as a bare OR chain, id = 1 or id = 2 or id = 3. The FDWs join quals with " and ", so:

SELECT count(*) FROM mysql_agg.orders WHERE id IN (1, 2, 3) AND status = 'inactive';

reaches the remote as

... where `id` = 1 or `id` = 2 or `id` = 3 and `status` = 'inactive'

AND binds tighter than OR, so ids 1 and 2 match whatever their status is.

• With aggregate pushdown (MySQL, ClickHouse, BigQuery), that wrong count or sum is returned directly: the query above returns 3 instead of 1.
• Plain scans are rechecked locally, so they only over-fetch.
• A plain scan still returns too few rows when a LIMIT is also pushed down.

The same pattern is in Qual::deparse (used by BigQuery, DuckDB, ClickHouse, MSSQL, and Wasm FDWs through the host's deparse) and in the copy in mysql_fdw::deparse_qual.

What is the new behavior?

An OR list with more than one element is wrapped in parentheses: (id = 1 or id = 2 or id = 3). Single-element and empty lists are rendered exactly as before.

Additional context

• New tests:
  • Unit tests in interface.rs: test_qual_deparse_or_list_is_parenthesized and test_qual_deparse_single_element_or_list.
  • A mysql_fdw case: COUNT(*) ... WHERE id IN (1, 2, 3) AND status = 'inactive' returns 1 and is still pushed down.
• Red/green:
  • On main, the unit test fails (left: "id = 1 or id = 2 or id = 3") and the mysql case fails with expected 1, got 3, run with cargo pgrx test pg17 mysql --features mysql_fdw against mysql:8.4.
  • With this change, cargo test -p supabase-wrappers --lib passes 38/38 and all 3 mysql pgrx tests pass.
• Checks: cargo fmt --check passes, and cargo clippy --tests -D warnings (pg17, mysql_fdw) is clean.

This PR was prepared with AI assistance (Claude). I reproduced the failure and ran the tests locally as described above.

🤖 Generated with Claude Code

[imor]

Closing as this is 100% AI generated.