Scriban has a Stack Overflow via Nested Array Initializers That Bypass the ExpressionDepthLimit Fix

Summary

### Summary StackOverflowException via nested array initializers bypasses ExpressionDepthLimit fix (GHSA-wgh7-7m3c-fx25) ### Details The recent fix for GHSA-wgh7-7m3c-fx25 (uncontrolled recursion in parser) added `ExpressionDepthLimit` defaulting to 250. However, deeply nested **array initializers** (`[[[[...`) recurse through `ParseArrayInitializer` → `ParseExpression` → `ParseArrayInitializer`, which is a **different recursion path** not covered by the expression depth counter. This causes a `StackOverflowException` on current main (commit b5ac4bf - "Add limits for default safety"). ### PoC ``` using Scriban; // ExpressionDepthLimit (default 250) does NOT prevent this crash string nested = "{{ " + new string('[', 5000) + "1" + new string(']', 5000) + " }}"; Template.Parse(nested); // StackOverflowException - process terminates ``` ### Impact Same as GHSA-wgh7-7m3c-fx25: High severity. StackOverflowException cannot be caught with try/catch in .NET - the process terminates immediately. Any application calling Template.Parse with untrusted input is vulnerable, even with the new default ExpressionDepthLimit enabled.