Coverity complained about possible buffer overrun in two places added by
commit 1eb6d652, and AFAICS it's reasonable to worry: even granting that
the WAL originator properly truncated the commit GID to GIDSIZE, we should
not really bet our lives on that having the same value as it does in the
current build.  Hence, use strlcpy() not strcpy(), and adjust the pointer
advancement logic to be sure we skip over the whole source string even if
strlcpy() truncated it.