Add support for PatternMatchMapping formatting (#6836)

<!--
Thank you for contributing to Ruff! To help us out with reviewing,
please consider the following:

- Does this pull request include a summary of the change? (See below.)
- Does this pull request include a descriptive title?
- Does this pull request include references to any relevant issues?
-->

## Summary

Adds support for `PatternMatchMapping` -- i.e., cases like:

```python
match foo:
    case {"a": 1, "b": 2, **rest}:
        pass
```

Unfortunately, this node has _three_ kinds of dangling comments:

```python
{  # "open parenthesis comment"
   key: pattern,
   **  # end-of-line "double star comment"
   # own-line "double star comment"
   rest  # end-of-line "after rest comment"
   # own-line "after rest comment"
}
```

Some of the complexity comes from the fact that in `**rest`, `rest` is
an _identifier_, not a node, so we have to handle comments _after_ it as
dangling on the enclosing node, rather than trailing on `**rest`. (We
could change the AST to use `PatternMatchAs` there, which would be more
permissive than the grammar but not totally crazy -- `PatternMatchAs` is
used elsewhere to mean "a single identifier".)

Closes https://github.com/astral-sh/ruff/issues/6644.

## Test Plan

`cargo test`

crates/ruff_python_formatter/src/pattern/pattern_match_mapping.rs

@@ -1,23 +1,103 @@
-use ruff_formatter::{write, Buffer, FormatResult};
+use ruff_formatter::{format_args, write};
 use ruff_python_ast::node::AnyNodeRef;
 use ruff_python_ast::PatternMatchMapping;
+use ruff_python_ast::{Expr, Identifier, Pattern, Ranged};
+use ruff_python_trivia::{SimpleTokenKind, SimpleTokenizer};
+use ruff_text_size::TextRange;
 
-use crate::expression::parentheses::{NeedsParentheses, OptionalParentheses};
+use crate::comments::{leading_comments, trailing_comments, SourceComment};
+use crate::expression::parentheses::{
+    empty_parenthesized, parenthesized, NeedsParentheses, OptionalParentheses,
+};
 use crate::prelude::*;
-use crate::{not_yet_implemented_custom_text, FormatNodeRule, PyFormatter};
 
 #[derive(Default)]
 pub struct FormatPatternMatchMapping;
 
 impl FormatNodeRule<PatternMatchMapping> for FormatPatternMatchMapping {
     fn fmt_fields(&self, item: &PatternMatchMapping, f: &mut PyFormatter) -> FormatResult<()> {
-        write!(
-            f,
-            [not_yet_implemented_custom_text(
-                "{\"NOT_YET_IMPLEMENTED_PatternMatchMapping\": _, 2: _}",
-                item
-            )]
-        )
+        let PatternMatchMapping {
+            keys,
+            patterns,
+            rest,
+            range: _,
+        } = item;
+
+        debug_assert_eq!(keys.len(), patterns.len());
+
+        let comments = f.context().comments().clone();
+        let dangling = comments.dangling(item);
+
+        if keys.is_empty() && rest.is_none() {
+            return empty_parenthesized("{", dangling, "}").fmt(f);
+        }
+
+        // This node supports three kinds of dangling comments. Most of the complexity originates
+        // with the rest pattern (`{**rest}`), since we can have comments around the `**`, but
+        // also, the `**rest` itself is not a node (it's an identifier), so comments that trail it
+        // are _also_ dangling.
+        //
+        // Specifically, we have these three sources of dangling comments:
+        // ```python
+        // {  # "open parenthesis comment"
+        //    key: pattern,
+        //    **  # end-of-line "double star comment"
+        //    # own-line "double star comment"
+        //    rest  # end-of-line "after rest comment"
+        //    # own-line "after rest comment"
+        // }
+        // ```
+        let (open_parenthesis_comments, double_star_comments, after_rest_comments) =
+            if let Some((double_star, rest)) = find_double_star(item, f.context().source()) {
+                let (open_parenthesis_comments, dangling) =
+                    dangling.split_at(dangling.partition_point(|comment| {
+                        comment.line_position().is_end_of_line()
+                            && comment.start() < double_star.start()
+                    }));
+                let (double_star_comments, after_rest_comments) = dangling
+                    .split_at(dangling.partition_point(|comment| comment.start() < rest.start()));
+                (
+                    open_parenthesis_comments,
+                    double_star_comments,
+                    after_rest_comments,
+                )
+            } else {
+                (dangling, [].as_slice(), [].as_slice())
+            };
+
+        let format_pairs = format_with(|f| {
+            let mut joiner = f.join_comma_separated(item.end());
+
+            for (key, pattern) in keys.iter().zip(patterns) {
+                let key_pattern_pair = KeyPatternPair { key, pattern };
+                joiner.entry(&key_pattern_pair, &key_pattern_pair);
+            }
+
+            if let Some(identifier) = rest {
+                let rest_pattern = RestPattern {
+                    identifier,
+                    comments: double_star_comments,
+                };
+                joiner.entry(&rest_pattern, &rest_pattern);
+            }
+
+            joiner.finish()?;
+
+            trailing_comments(after_rest_comments).fmt(f)
+        });
+
+        parenthesized("{", &format_pairs, "}")
+            .with_dangling_comments(open_parenthesis_comments)
+            .fmt(f)
+    }
+
+    fn fmt_dangling_comments(
+        &self,
+        _dangling_comments: &[SourceComment],
+        _f: &mut PyFormatter,
+    ) -> FormatResult<()> {
+        // Handled by `fmt_fields`
+        Ok(())
     }
 }
 
@@ -30,3 +110,78 @@ impl NeedsParentheses for PatternMatchMapping {
         OptionalParentheses::Never
     }
 }
+
+/// A struct to format the `rest` element of a [`PatternMatchMapping`] (e.g., `{**rest}`).
+#[derive(Debug)]
+struct RestPattern<'a> {
+    identifier: &'a Identifier,
+    comments: &'a [SourceComment],
+}
+
+impl Ranged for RestPattern<'_> {
+    fn range(&self) -> TextRange {
+        self.identifier.range()
+    }
+}
+
+impl Format<PyFormatContext<'_>> for RestPattern<'_> {
+    fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
+        write!(
+            f,
+            [
+                leading_comments(self.comments),
+                text("**"),
+                self.identifier.format()
+            ]
+        )
+    }
+}
+
+/// A struct to format a key-pattern pair of a [`PatternMatchMapping`] (e.g., `{key: pattern}`).
+#[derive(Debug)]
+struct KeyPatternPair<'a> {
+    key: &'a Expr,
+    pattern: &'a Pattern,
+}
+
+impl Ranged for KeyPatternPair<'_> {
+    fn range(&self) -> TextRange {
+        TextRange::new(self.key.start(), self.pattern.end())
+    }
+}
+
+impl Format<PyFormatContext<'_>> for KeyPatternPair<'_> {
+    fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
+        write!(
+            f,
+            [group(&format_args![
+                self.key.format(),
+                text(":"),
+                space(),
+                self.pattern.format()
+            ])]
+        )
+    }
+}
+
+/// Given a [`PatternMatchMapping`], finds the range of the `**` element in the `rest` pattern,
+/// if it exists.
+fn find_double_star(pattern: &PatternMatchMapping, source: &str) -> Option<(TextRange, TextRange)> {
+    let PatternMatchMapping {
+        keys: _,
+        patterns,
+        rest,
+        range: _,
+    } = pattern;
+
+    // If there's no `rest` element, there's no `**`.
+    let Some(rest) = rest else {
+        return None;
+    };
+
+    let mut tokenizer =
+        SimpleTokenizer::starts_at(patterns.last().map_or(pattern.start(), Ranged::end), source);
+    let double_star = tokenizer.find(|token| token.kind() == SimpleTokenKind::DoubleStar)?;
+
+    Some((double_star.range(), rest.range()))
+}