Preserve and format type aliases in extern blocks

[ayazhafiz]

Previously, non-trivial type aliases in extern blocks were dropped by
rustfmt because only the type alias name would be passed to a rewritter.
This commit fixes that by passing all type information (generics,
bounds, and assignments) to a type alias rewritter, and consolidates
rewrite_type_alias and rewrite_associated_type as one function.

Closes #4159

[calebcartwright]

Do we always have to add a newline and/or is this something we'd expect users to want to configure? I get the sense that this will be fairly subjective and some folks won't be fond of all the multilining for these type aliases

[ayazhafiz]

I copied this from the existing implementation for writing types:

rustfmt/rustfmt-core/rustfmt-lib/src/items.rs

Lines 1599 to 1623 in 12564f1

	fn rewrite_type_item<R: Rewrite>(
	context: &RewriteContext<'_>,
	indent: Indent,
	prefix: &str,
	suffix: &str,
	ident: ast::Ident,
	rhs: &R,
	generics: &ast::Generics,
	vis: &ast::Visibility,
	) -> Option<String> {
	let mut result = String::with_capacity(128);
	result.push_str(&rewrite_type_prefix(
	context,
	indent,
	&format!("{}{} ", format_visibility(context, vis), prefix),
	ident,
	generics,
	)?);
	if generics.where_clause.predicates.is_empty() {
	result.push_str(suffix);
	} else {
	result.push_str(&indent.to_string_with_newline(context.config));
	result.push_str(suffix.trim_start());
	}

[calebcartwright]

Yeah I figured, and I know it's consistent (both with existing type alias formatting and on functions/structs/enums). I suppose folks can also use where_single_line to control things a bit as well.

It just felt odd to me that an alias as short as type A<'a> where 'a: 'static; would be split into 4 lines by default (understood that will become 3 after the semi placement change discussed above)

[ayazhafiz]

Yeah, I agree it looks a little odd. I think keeping consistency with how where clauses are formatted in other contexts (i.e. writing the where on a newline) is valuable. Do you find the 3-line version with the semicolon placement change to be nicer?

[topecongiro]

Yeah, ideally, we respect the where_single_line option wherever we format a where clause. Unfortunately, that's not the case in the current implementation. I think that we can fix this as a part of the stabilization process of the where_single_line option; in this PR, we can follow the existing approach.

[calebcartwright]

SGTM 👍

[ayazhafiz]

Sorry this diff is not the prettiest, it may be easier to look at the raw file.

[topecongiro]

LGTM, thank you for the update!

[karyon]

backported in #4447