batch_query result count misreports filtered errors as executed queries

[Widthdom]

Summary

ExecuteBatchQuery at /Users/widthdom/Projects/mine/cdidx/CodeIndex/src/CodeIndex/Mcp/McpToolHandlers.cs:1215 returns resultsArray.Count in the summary, but this conflates errors (blocked write operations, missing tool names, validation failures) with actual executed queries. A batch with 10 queries where 8 fail validation will still report "Executed 10 queries" even though only 2 actually ran. This misleads AI clients about batch completion scope and success rate.

Where

• McpToolHandlers.cs:1127-1219
• McpToolHandlers.cs:1215 result count aggregation

Suggested approach

1. Split resultsArray counting into successCount and errorCount tracks during loop.
2. Track actuallyExecuted separately from totalSubmitted.
3. Return both counts in the payload: ["submitted"] = queries.Count, ["executed"] = actuallyExecuted, ["errors"] = errorCount.
4. Update summary text to report "Executed X of Y queries (Z errors)" format.
5. Add fixture tests covering 100% error batches, partial-error batches, and all-success batches.
6. Document in tool definition that count reports actual executions, not input batch size.

[Widthdom]

Work started on branch codex/fix-issue1838-1992-1994-2028.