Fix string offset parameters to use character positions instead of byte offsets

[Claude]

The Find(string, int) and Captures(string, int) methods were treating the offset parameter as a byte position in the UTF-8 encoded string, causing incorrect behavior with multi-byte characters (emojis, CJK characters, etc.).

Changes

• Modified Find(string haystack, int offset) and Captures(string haystack, int offset): Convert character offset to byte offset using Encoding.UTF8.GetByteCount(haystack.AsSpan(0, offset)) before passing to the underlying byte-based methods
• Updated XML documentation: Clarified that string-based methods accept character offsets while byte-based methods accept byte offsets
• Added test coverage: 6 new tests covering multi-byte UTF-8 scenarios (4-byte emojis, 3-byte CJK characters)

Example

Regex re = new("world");
string haystack = "Hello 🌍 world";  // 🌍 is 4 bytes in UTF-8

// Character positions: H(0) e(1) l(2) l(3) o(4) (5) 🌍(6) (7) w(8)...
// Byte positions: H(0) e(1) l(2) l(3) o(4) (5) 🌍(6-9) (10) w(11)...

// Now correctly interprets offset as character position
var match = re.Find(haystack, 8);  // Starts at character 'w', not mid-emoji
Assert.Equal("world", match.ExtractedText);