fix(statusline): "%.0" renders garbage #41685

Problem:
`getdigits_int(pp, strict, def)` returns `def` only when the number
fails to parse or overflows, never when the text parses to a literal
`0`, so `%.0` leaves `maxwid` at 0. Both render paths then break down:
`l > maxwid` is always true and `while (l >= maxwid)` consumes the whole
string, so `%.0f` comes out as `<` and `%.0l` as `0>3`.

Vim guards for this right after the same call. Nvim had the guard too
until 3344cffe7b ("getdigits: introduce `strict`, `def` parameters"),
which folded it into the new `def` argument and so changed what a
literal zero means. The neighbouring `minwid` call was translated
correctly.

Solution:
Fall back to 50 when the parsed width is not positive, as Vim does.

AI-assisted

src/nvim/statusline.c

@@ -1377,6 +1377,9 @@ int build_stl_str_hl(win_T *wp, char *out, size_t outlen, char *fmt, OptIndex op
       fmt_p++;
       if (ascii_isdigit(*fmt_p)) {
         maxwid = getdigits_int(&fmt_p, false, 50);
+        if (maxwid <= 0) {  // overflow or an explicit zero
+          maxwid = 50;
+        }
       }
     }
 

test/functional/api/vim_spec.lua

@@ -4537,6 +4537,13 @@ describe('API', function()
       }, api.nvim_eval_statusline('Should be truncated%<', { maxwidth = 15 }))
     end)
 
+    it('does not take a zero item width literally', function()
+      command('file some/dir/testfile.txt')
+      eq({ str = 'abc', width = 3 }, api.nvim_eval_statusline('%.0(abc%)', {}))
+      eq(api.nvim_eval_statusline('%.50f', {}), api.nvim_eval_statusline('%.0f', {}))
+      eq(api.nvim_eval_statusline('%.50l', {}), api.nvim_eval_statusline('%.0l', {}))
+    end)
+
     it('has correct default fillchar', function()
       local oldwin = api.nvim_get_current_win()
       command('set fillchars=stl:#,stlnc:$,wbr:%')