[Nickle] parse-args changes, command-line revision

[nickle at po8.org]

I've completely revised the Nickle command-line argument
parsing code.  It now uses the Nickle parse-args stuff
instead of ad hoc for argument parsing.  In the process,
I've changed the Nickle command line in some subtle ways:
   * The -e flag is now limited to evaluating a single
     argument.
   * You can now have an arbitrary sequence of -e, -f, and -l
     flags that will be executed in the order given.
   * The -l flag and -f flag can now be used when starting
     an interactive session.
   * You can now get a non-empty argv in interactive mode.  Use
     the argument-parse-stop argument "--" and give any
     arguments you want after that.
   * You can now get sensible command-line help with --help
     or --usage.
   * Arguments are now checked much more carefully.   

In the process, I made some substantial incompatible changes
to parse-args.5c.
   * Positional (non-flagged) arguments are now handled with
     a separate array rather than the previous kludge in
     with the flags.
   * Positional arguments may now be optional.
   * If the program name is preset in the argd structure,
     argv[0] is presumed to be the first argument, not
     the program name.  (Needed for Nickle startup).
   * There is now an option for getting the index of
     the first argument unconsumed by parseargs().
   * There is now an argument type of arg_lambda, which
     allows supplied code to be executed as arguments
     are parsed.
If you were using ParseArgs::parseargs(), you may have to
rewrite your argument descriptions a bit.

Comments and bug reports welcome.

	Bart