update_view_title: Fix NULL dereference at startup

update_view_title can be called with a `struct view`
where `line` is NULL, and `lines` is 0.

Specifically, along this call stack:

    0  update_view_title (view=0x3fdb88 <main_view>) at [...]/tig/src/view.c:690
    1  0x0000000000338018 in report_clear () at [...]/tig/src/display.c:565
    2  0x00000000003cfe5b in load_view (view=0x3fdb88 <main_view>, prev=0x3fdb88 <main_view>, flags=OPEN_
       at [...]/tig/src/view.c:857
    3  0x00000000003d0bc0 in open_view (prev=0x0, view=0x3fdb88 <main_view>, flags=OPEN_DEFAULT)
       at [...]/tig/src/view.c:894
    4  0x00000000003b2932 in open_main_view (prev=0x0, flags=OPEN_DEFAULT) at include/tig/main.h:57
    5  0x00000000003b0cca in view_driver (view=0x0, request=REQ_VIEW_MAIN) at [...]/tig/src/tig.c:179
    6  0x00000000003af96a in main (argc=1, argv=0x7fffffffddb8) at [...]/tig/src/tig.c:864

Specifically, load_view calls report_clear when `view->lines == 0`,
which calls `update_view_title`, which attempts `&view->line[...]`
on a null `line`.

It's not clear why this doesn't explode today.
I caught it when I ran tig compiled with Zig in debug mode
and it failed with an illegal instruction on the line:

    struct line *line = &view->line[view->pos.lineno];

Adding a check for `NULL` or `lines == 0` resolves the issue.

src/view.c

@@ -687,6 +687,8 @@ void
 update_view_title(struct view *view)
 {
 	WINDOW *window = view->title;
+	if (view->line == NULL)
+		return;
 	struct line *line = &view->line[view->pos.lineno];
 	unsigned int view_lines, lines;
 	int update_increment = view_has_flags(view, VIEW_LOG_LIKE | VIEW_GREP_LIKE)