Skip over CRs at the end of multiline literals

Fixes ziglang#9257.
This is needed when tokenizing input containing DOS line endings, i.e.
the CRLF sequence.

Author: jmc-88

lib/std/zig/tokenizer.zig

@@ -834,7 +834,7 @@ pub const Tokenizer = struct {
                 },
 
                 .multiline_string_literal_line => switch (c) {
-                    '\n' => {
+                    '\r', '\n' => {
                         self.index += 1;
                         break;
                     },
@@ -2049,6 +2049,10 @@ test "tokenizer - invalid builtin identifiers" {
     try testTokenize("@0()", &.{ .invalid, .integer_literal, .l_paren, .r_paren });
 }
 
+test "tokenizer - multiline string literal with a DOS-encoded CRLF line endings (issue #9257)" {
+    try testTokenize("\\\\foobar\r\n", &.{.multiline_string_literal_line});
+}
+
 fn testTokenize(source: []const u8, expected_tokens: []const Token.Tag) !void {
     var tokenizer = Tokenizer.init(source);
     for (expected_tokens) |expected_token_id| {