Fix chop_str not to cut in middle of utf8 multibyte chars. Without
this fix at least author name in short log may cut in middle of a
multibyte char. When the result comes to esc_html to_utf8 is called
again, which doesn't find valid utf8 and decodes using
$fallback_encoding making it even worse.

This also have the nice side effect that it actually tries to show the
first 10 _characters_, not the number of characters that happened to fit
into 10 bytes.
Signed-off-by: NAnders Waldenborg <anders@0x63.nu>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>