Scroll to follow expanding part of editor::SelectLargerSyntaxNode (#27295)

When the selection grows both ways, the new code prioritizes the top
part instead of bottom one, this is usually more helpful considering
that most programming language grammars tend to define tokens right
before large delimited blocks, and rarely after (because humans and
parsers read from top to bottom).

Also, revert selection when convenient, so you have more control over
what you're selecting, looking at the selection `head` is commonly more
convenient than at the `tail`.

Release Notes:

- Improve scrolling of `editor::SelectLargerSyntaxNode` for better
visibility.

crates/editor/src/editor_tests.rs

@@ -6033,7 +6033,7 @@ async fn test_select_larger_smaller_syntax_node(cx: &mut TestAppContext) {
                 use mod1::mod2::{mod3, «mod4ˇ»};
 
                 fn fn_1«ˇ(param1: bool, param2: &str)» {
-                    let var1 = "«textˇ»";
+                    let var1 = "«ˇtext»";
                 }
             "#},
             cx,
@@ -6101,7 +6101,7 @@ async fn test_select_larger_smaller_syntax_node(cx: &mut TestAppContext) {
                 use mod1::mod2::{mod3, «mod4ˇ»};
 
                 fn fn_1«ˇ(param1: bool, param2: &str)» {
-                    let var1 = "«textˇ»";
+                    let var1 = "«ˇtext»";
                 }
             "#},
             cx,
@@ -6170,7 +6170,7 @@ async fn test_select_larger_smaller_syntax_node(cx: &mut TestAppContext) {
                 use mod1::mod2::«{mod3, mod4}ˇ»;
 
                 fn fn_1«ˇ(param1: bool, param2: &str)» {
-                    «let var1 = "text";ˇ»
+                    «ˇlet var1 = "text";»
                 }
             "#},
             cx,