send.c:_pipecmd(): fix [81473f8]: again ensure *result is set..

send.c:_pipecmd(): fix [81473f8]: again ensure *result is set..

In [81473f8] (Add *pipe-EXTENSION*, ext. *mime-counter-evidence*..,
2014-07-11) i unfortunately broke the special `@' text handlers by
no longer setting *result to NULL at function entry.

The real cause of this bug however is the temporary colourization
hack (i don't want to miss it again), since when i implemented the
extended *mime-counter-evidence* stuff i reduced _pipecmd() to
it's essence, the problem being that the local variable
send.c:sendpart():pipecomm gets misused as temporary storage for
the colourization -- due to this it is *not* NULL if _pipecmd()
returns PIPE_TEXT to indicate forced treatment as text.

I haven't yet seen this rather ultimate MCA because i do ignore
all header fields (which come into play for MIME parts) for `p',
and only display them for `P'.
v14.8 will come with a test series.  (sigh.)