[ty] Do not suggest argument completion when at value of keyword argument (#24669)

In the following situation:

```python
def foo(y_true,y_pred): ...

y_true = 1
y_pred = 2

foo(y_true=y<CURSOR>)
```

the completion suggestions began with `y_true=` and `y_pred=`. But it is
never the right thing to suggest an argument completion (i.e.
`something=`) when the cursor is at the _value_ of a keyword argument.
So this PR introduces an early exit to the logic for deciding to suggest
arguments.

Author: dylwil3

crates/ty_ide/src/completion.rs

@@ -1411,6 +1411,13 @@ fn add_argument_completions<'db>(
     let mut in_arguments = false;
     for node in cursor.covering_node.ancestors() {
         match node {
+            // Do not suggest argument completions in value positions for
+            // keyword arguments
+            ast::AnyNodeRef::Keyword(kw) => {
+                if kw.value.range().contains_range(cursor.range) {
+                    return;
+                }
+            }
             ast::AnyNodeRef::Arguments(_) => {
                 in_arguments = true;
             }
@@ -8733,6 +8740,28 @@ re.match('', '', fla<CURSOR>
         );
     }
 
+    #[test]
+    fn call_keyword_argument_at_value() {
+        let builder = completion_test_builder(
+            "\
+def bar(y_true,y_pred): ...
+
+y_true = 1
+y_pred = 2
+
+bar(y_true=y<CURSOR>
+",
+        );
+
+        assert_snapshot!(
+            builder.skip_keywords().skip_builtins().skip_auto_import().build().snapshot(),
+            @r###"
+        y_pred
+        y_true
+        "###
+        );
+    }
+
     // Ideally, we should favour completions that are definitely raisable
     // here. However, doing so would require `exception_ty` to fall back to
     // token matching when AST-matching fails, making the function signficantly