Add token limit handling to prevent 8k token overflow

[YosefHayim]

This PR adds token limit handling to prevent API failures when large git diffs exceed the 8k token limit.

Problem

• GitHub Models API has an 8k token limit for entire requests
• Large git diffs can exceed this limit, causing API failures
• Users experience failures when staging large changes

Solution

• Added token estimation using character-based heuristic (1 token ≈ 4 chars)
• Implemented truncation logic that preserves UTF-8 boundaries
• Added intelligent content prioritization when over limit

Implementation Details

• estimateTokens(): Approximates tokens for any text content
• truncateToTokenLimit(): Safely truncates text with ellipsis indicator
• Modified GenerateCommitMessage() to:
  • Estimate tokens for prompt templates + changes + examples
  • Reserve tokens for templates (with buffer)
  • Prioritize examples (20% of remaining tokens) when present
  • Truncate changes to fit remaining budget
  • Display warning when truncation occurs

Benefits

• Prevents API failures from token overflow
• Maintains functionality by preserving maximum content
• User-friendly with clear truncation warnings
• No external dependencies, follows existing code style
• Gracefully handles both changes-only and changes+examples scenarios

Testing

• Code compiles successfully
• Follows existing project patterns and style
• Token estimation uses common industry heuristic