Use edit-distance alignment for accuracy / character error stats

[Ramji-Purwar]

Current character accuracy appears to compare input and target strictly by position. This causes one inserted or missing character to shift alignment, marking all following characters as incorrect.
Example:

Target: Hello
Typed: hHello

The typed text differs by one extra h (Levenshtein distance = 1), but a position-based comparison can count every subsequent character as wrong.

Could accuracy and character stats use an alignment-aware edit-distance calculation instead? This would count insertions, deletions, and substitutions as localized errors, rather than cascading a single mistake across the rest of the word.