Switch buffer from bytes to characters for proper multibyte handling

[tk0miya]

Summary

This PR refactors the code rendering system to work with character positions instead of byte positions. This fixes incorrect handling of multibyte characters (e.g., emoji, CJK characters) in ERB templates.

Key Changes

• Buffer representation: Changed from Array[Integer] (bytes) to Array[String] (characters)

  • Simplifies position calculations and eliminates byte/character conversion complexity
  • Final code generation now uses buffer.join instead of pack("C*")

• Character constants: Added string equivalents for all byte constants in Characters module

  • CHAR_LF, CHAR_CR, CHAR_SPACE, CHAR_HASH, CHAR_SEMICOLON, CHAR_EQUALS, CHAR_UNDERSCORE, CHAR_LEFT_BRACE, CHAR_RIGHT_BRACE
  • Enables direct character manipulation without byte conversion

• Position conversion: Implemented byte_to_char_pos method in Source class

  • Converts Herb's byte-based positions to character positions for RuboCop compatibility
  • Delegated through ParseResult and RubyRenderer

• Converter refactoring: Updated generate_hybrid_code to work with character arrays

  • Uses byte_to_char_pos to convert positions before buffer manipulation
  • Properly handles multibyte character restoration

• RubyRenderer updates: All position calculations now use character positions

  • render_code_node, render_output_marker, render_open_tag_node, render_close_tag_node, render_text_node, render_erb_comment_node, render_html_comment_node, render_tag_marker all updated
  • bleach_code now returns Array[String] and uses character-based logic

• Test updates: Fixed test expectations to use character counts instead of byte counts

  • Multibyte character tests now show correct spacing (e.g., "日本語" is 3 characters, not 9 bytes)

Implementation Details

The core insight is that RuboCop operates on character positions, while Herb's parser provides byte positions. By converting the buffer to character-based operations, we eliminate the mismatch that caused incorrect positioning with multibyte characters. The byte_to_char_pos conversion is applied at the boundary where Herb positions are used in RuboCop's buffer.