fix: handle square bracket HTML entities in Gemini responses

[roomote]

Description

This PR fixes an issue where square brackets and array indices were being stripped from C# code when using Google Gemini models (specifically gemini-2.5-pro).

Problem

When using Gemini API with Roo Code to generate C# code containing arrays, the square brackets were missing from the output. For example:

• Expected: string[] myArray = new string[5];
• Actual: string myArray = new string5;

Solution

The issue was that Gemini was escaping square brackets as HTML entities ([, ], [, ]), but our unescapeHtmlEntities function wasn't handling these specific entities.

This PR adds support for unescaping both numeric and named HTML entities for square brackets in the unescapeHtmlEntities function.

Changes

• Added support for [, ], [, and ] entities in src/utils/text-normalization.ts
• Added comprehensive tests for square bracket entity handling
• Tests include the exact scenario from the bug report (C# array syntax)

Testing

• All existing tests pass ✅
• New tests added for square bracket entity handling ✅
• Tested with the specific C# array syntax from the issue ✅

Fixes #7576

---

Important

Fixes missing square brackets in C# code by updating unescapeHtmlEntities to handle specific HTML entities.

• Behavior:
  • Fixes issue with missing square brackets in C# code generated by Gemini API by updating unescapeHtmlEntities in text-normalization.ts.
  • Supports unescaping [, ], [, and ] entities.
• Testing:
  • Adds tests for numeric and named square bracket entities in text-normalization.spec.ts.
  • Includes test for C# array syntax with escaped square brackets.

^{This description was created by for 5b2fd3f. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Reviewing my own code because apparently I trust no one, not even myself.

[roomote]

Is the order of replacements intentional here? I notice that the ampersand replacement is placed last, which seems correct since replacing it first could interfere with other entity replacements. Would it be worth adding a comment explaining why it must be last to prevent future reordering mistakes?

[roomote]

While this PR fixes the square bracket issue, are there other HTML entities that Gemini might escape that we're not handling? For example: parentheses (( and )), curly braces ({ and }), or other mathematical/programming symbols. Might be worth investigating what other entities Gemini commonly escapes in code contexts.

[roomote]

Would it make sense to group all the square bracket tests together in a nested describe block for better organization? This would make the test structure clearer and easier to navigate.

[roomote]

Great test coverage! This test case directly addresses the reported issue with C# array syntax. The test is well-documented with a comment explaining its purpose.

[daniel-lxs]

LGTM