A bunch of the lingering EasyListening reactors are doing that because
they want commands like "verb subthing" and "verb othersubthing".  The
CommandPost translations of that would come in two forms, until now,
neither very appealing.

For one, you might make a "verb" command which then subparses.  This
loses a bunch of nice properties of commandpost, like the parser system.

For the other, you could make a responder, which just has the same
problem.

I've long considered a means to say "add this reactor with a prefix
word", so you could make "subthing" and "othersubthing" commands, and
give the reactor a "verb" prefix.  That would also allow the reactor to
be configured twice with distinct prefixes.  I think this is still a
good idea.

The change made in this commit, though, is smaller in scope.  It allows
command names and aliases to have spaces in them.  If they do, they're
matched differently.  We can't just split off the first word and look
things up in a hash.  Instead, we do a regex match.  To avoid doing 100
regex matches per message if there are 100 commands, we *only* do the
regex match for multi-word commands, and only if any multi-word commands
have been registered.  This implementation detail should be invisible to
the user, and so can be changed later.

Prefix conflicts (the command "eat" and "eat pie" both being registered,
for example) are not handled.  They will handled implicitly by the fact
that the PotentialReactions of commands are always flagged exclusive.