Feature: Add support for stripping XML comments and declarations

[GunwooYun]

Description

This PR implements a pre-processing step in xml_parse_document to strip XML comments (<!-- ... -->) and XML declarations (<? ... ?>) from the input buffer before parsing.

Prior to this change, the parser would likely choke on comments or declarations as it expects a strictly structured validation-ready XML or a specific subset. This change allows the parser to handle more standard XML files that include comments or headers.

Implementation Details

• Modified xml_parse_document in src/xml.c.
• Implemented an in-place buffer modification loop that shifts valid bytes to overwrite comments and declarations.
• Updated document_length to reflect the new, shorter content length.

Testing

I have created a test suite in app/ including a custom test runner app/test_runner.c and several test cases:

• Basic Comments:

  <!-- comment -->
  <root>content</root>

• Multiline Comments:

  <!-- 
   multi 
   line 
   -->
  <root>content</root>

• XML Declarations:

  <?xml version="1.0"?>
  <root>content</root>

• Nested Comments:

  <root>
      <level1>
          <!-- level 1 comment -->
          <level2>...</level2>
      </level1>
      <!-- sibling comment -->
  </root>

• Consecutive Comments:

  <!-- 1 --><!-- 2 -->
  <root>...</root>

All the above tests pass successfully.

Known Limitations

The current implementation is a naive approach. It blindly searches for <!-- and --> markers without context.

• Tricky Attributes: If a comment marker appears inside an attribute value (e.g., <node attr="<!--">), it will be incorrectly identified as a comment start, corrupting the XML.
• Unclosed Comments: An unclosed comment might be interpreted as a node named !-- or cause malformed output, as it doesn't strictly validate comment syntax.

I have added a warning comment in the code highlighting these limitations. For the intended lightweight use case of this library, this trade-off was chosen for simplicity and performance (O(n) single pass).