Pg-Meta IS Filter Fails To Normalize Uppercase SQL Keywords

`isFilterSql` in `packages/pg-meta/src/query/Query.utils.ts` compares the raw filter value directly against lowercase literals (`'null'`, `'true'`, `'false'`, `'not null'`). Any uppercase or mixed-case input fails the comparison, falls through to the quoted-string branch, and produces SQL like `where email is 'NULL'` instead of `where email is null`.

1. Use the pg-meta query builder with `.filter('email', 'is', 'NULL')`; 2. Call `.toSql()` to generate the SQL string; 3. Observe that the output is `select * from public.users where email is 'NULL';` instead of `select * from public.users where email is null;`. The same issue occurs for 'NOT NULL', 'TRUE', and 'FALSE'.

This change could theoretically affect users who intentionally filter for the literal string 'NULL' (lowercase) or 'null' using the `is` operator; such queries would still be interpreted as SQL NULL, though that behavior already existed for lowercase input. If the value is not a string (e.g., numeric or boolean), `String(value)` will coerce it, but that matches typical usage. The fix is localized and unlikely to affect concurrency, multi-threading, or cross-platform behavior since it is a pure string transformation. Rollback is straightforward: revert to the previous case-sensitive comparison. To avoid future regressions, consider centralizing keyword normalization in a shared utility and adding unit tests for all keyword spellings.