read: Don't use chunking read with --line

Fixes a regression from #8552.

src/builtins/read.cpp

@@ -501,6 +501,10 @@ maybe_t<int> builtin_read(parser_t &parser, io_streams_t &streams, const wchar_t
                 read_interactive(parser, buff, opts.nchars, opts.shell, opts.silent, opts.prompt,
                                  opts.right_prompt, opts.commandline, streams.stdin_fd);
         } else if (!opts.nchars && !stream_stdin_is_a_tty &&
+                   // "one_line" is implemented as reading n-times to a new line,
+                   // if we're chunking we could get multiple lines so we would have to advance
+                   // more than 1 per run through the loop. Let's skip that for now.
+                   !opts.one_line &&
                    (streams.stdin_is_directly_redirected || lseek(streams.stdin_fd, 0, SEEK_CUR) != -1)) {
             // We read in chunks when we either can seek (so we put the bytes back),
             // or we have the bytes to ourselves (because it's directly redirected).

tests/checks/read.fish

@@ -375,3 +375,10 @@ echo foo\nbar\nbaz | begin
     echo $bar
     # CHECK: bar
 end
+
+begin
+    echo 1
+    echo 2
+end | read -l --line foo bar
+echo $foo $bar
+# CHECK: 1 2