fix(plpgsql-deparser): fix PERFORM, INTO, and recfield bugs

[pyramation]

Summary

This PR fixes three bugs in the plpgsql-deparser that were causing invalid SQL output:

1. PERFORM SELECT fix: The PL/pgSQL parser stores PERFORM statements internally as "SELECT ...", so the deparser was outputting invalid PERFORM SELECT ... syntax. Now strips the leading SELECT keyword.

2. INTO clause depth-aware scanner: The old regex-based approach for inserting INTO clauses would incorrectly insert INTO inside subqueries. Replaced with a depth-aware scanner (~150 lines) that tracks parentheses, quotes, comments, and dollar quotes to find the correct insertion point only at depth 0.

3. Record field qualification: The deparser was returning just fieldname instead of the full qualified reference like new.is_active. Now uses recparentno to look up the parent record and construct the full reference.

Updates since last revision

• Whitespace normalization fix: The parser strips INTO <target> from the stored query but leaves the original whitespace behind. This was causing large gaps in output like SELECT x INTO y FROM z. Now normalizes leading whitespace after INTO insertion to a single space.

• Round-trip AST testing: Upgraded all 17 test cases to use round-trip testing (parse → deparse → reparse → compare cleaned ASTs). Added normalizeQueryWhitespace to cleanPlpgsqlTree to handle indentation differences in embedded SQL queries while preserving content inside string literals and dollar-quoted strings.

Review & Testing Checklist for Human

• Verify depth-aware scanner edge cases: The findIntoInsertionPoint function (lines 1090-1243) is complex state machine logic. Test with: nested subqueries, CTEs, UNION, dollar-quoted strings containing "FROM", comments containing keywords, escaped quotes
• Test generated SQL against PostgreSQL: Round-trip tests verify AST equality but don't validate the SQL is actually executable. Run a few test cases against a real PostgreSQL instance
• Review normalizeQueryWhitespace function: This function (~100 lines in test-utils/index.ts) normalizes whitespace for AST comparison. Verify it doesn't mask real semantic differences by being too aggressive
• Verify record field qualification in triggers: Test with OLD/NEW references in various trigger scenarios

Recommended test plan:

1. Run the new snapshot tests locally
2. Take a few of the snapshot outputs and execute them as PL/pgSQL function bodies in a real PostgreSQL database to verify syntax validity
3. Create a trigger function using OLD.field and NEW.field references and verify it parses/deparses correctly

Notes

This PR upstreams patches 1-3 from constructive-db PR #229. Patch 4 (RETURN statement handling) is being handled separately as it requires design discussion about passing return-type context.

Link to Devin run: https://app.devin.ai/sessions/8e1c971e9b194cd9a7dda034c89bd74b
Requested by: Dan Lynch (@pyramation)

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring