Lexer start of line is false only for Mode::Expression (#8880)

## Summary

This PR fixes the bug in the lexer where the `Mode::Ipython` wasn't
being considered when initializing the soft keyword transformer which
wraps the lexer. This means that if the source code starts with either
`match` or `type` keyword, then the keywords were being considered as
name tokens instead. For example,

```python
match foo:
    case bar:
        pass
```

This would transform the `match` keyword into an identifier if the mode
is `Ipython`.

The fix is to reverse the condition in the soft keyword initializer so
that any new modes are by default considered as the lexer being at start
of line.

## Test Plan

Add a new test case for `Mode::Ipython` and verify the snapshot.

fixes: #8870

crates/ruff_python_parser/src/snapshots/ruff_python_parser__lexer__tests__match_softkeyword_in_notebook.snap

@@ -0,0 +1,66 @@
+---
+source: crates/ruff_python_parser/src/lexer.rs
+expression: lex_jupyter_source(source)
+---
+[
+    (
+        Match,
+        0..5,
+    ),
+    (
+        Name {
+            name: "foo",
+        },
+        6..9,
+    ),
+    (
+        Colon,
+        9..10,
+    ),
+    (
+        Newline,
+        10..11,
+    ),
+    (
+        Indent,
+        11..15,
+    ),
+    (
+        Case,
+        15..19,
+    ),
+    (
+        Name {
+            name: "bar",
+        },
+        20..23,
+    ),
+    (
+        Colon,
+        23..24,
+    ),
+    (
+        Newline,
+        24..25,
+    ),
+    (
+        Indent,
+        25..33,
+    ),
+    (
+        Pass,
+        33..37,
+    ),
+    (
+        Newline,
+        37..37,
+    ),
+    (
+        Dedent,
+        37..37,
+    ),
+    (
+        Dedent,
+        37..37,
+    ),
+]