Remove lexer-based comment range detection (#5785)

## Summary

I'm doing some unrelated profiling, and I noticed that this method is
actually measurable on the CPython benchmark -- it's > 1% of execution
time. We don't need to lex here, we already know the ranges of all
comments, so we can just do a simple binary search for overlap, which
brings the method down to 0%.

## Test Plan

`cargo test`

crates/ruff_python_ast/src/source_code/comment_ranges.rs

@@ -10,6 +10,23 @@ pub struct CommentRanges {
     raw: Vec<TextRange>,
 }
 
+impl CommentRanges {
+    /// Returns `true` if the given range includes a comment.
+    pub fn intersects(&self, target: TextRange) -> bool {
+        self.raw
+            .binary_search_by(|range| {
+                if target.contains_range(*range) {
+                    std::cmp::Ordering::Equal
+                } else if range.end() < target.start() {
+                    std::cmp::Ordering::Less
+                } else {
+                    std::cmp::Ordering::Greater
+                }
+            })
+            .is_ok()
+    }
+}
+
 impl Deref for CommentRanges {
     type Target = [TextRange];
 

crates/ruff_python_ast/src/source_code/indexer.rs

@@ -93,7 +93,7 @@ impl Indexer {
     }
 
     /// Returns the byte offset ranges of comments
-    pub fn comment_ranges(&self) -> &CommentRanges {
+    pub const fn comment_ranges(&self) -> &CommentRanges {
         &self.comment_ranges
     }