day 20 fix POSIX handling

day 20 fix POSIX handling

I ought to fix my POSIX line split algorithm to explicitly fail if I
chunk it too small to actually contain the newline replacement for a
long line (or better yet, to auto-adjust without having to remember to
alter the chunk size each time I copy the code).  Doing that would be
easier if I moved my chew() macro into common.m4.  But for now, just
fix the immediate 2 problems: chunking a line too small can result in
calling do(`') which inf-loops, and an input file might have a chunk
end in 'broad' which triggers an unexpected broadcast.