We use Buffer.Node.get_byte_pos instead of Buffer.Node.get_range to calculate tree-sitter edits for incremental parsing. This reduces the number of egc_length calls required per syntax.edit call to the minimum.
Metrix was a bit too silly