Multiport iterator for more efficient access to sparse inputs

[edwardalee]

This PR enables more efficient access to input multiports that are sparse, meaning that few of the channels (typically) have inputs present. This is used as in the following example:

     @sparse
     input[100] in:int;
     reaction(in) {=
         struct lf_multiport_iterator_t i = lf_multiport_iterator(in);
         int channel = lf_multiport_next(&i);
         while(channel >= 0) {
             printf("Received %d on channel %d.\n", in[channel]->value, channel);
             channel = lf_multiport_next(&i);
         }
     =}

In my informal measurement, this yields a 15% improvement in the unthreaded execution of the Big benchmark (with the default setting of 120 Worker reactors).

This PR also includes fixes to the annotation scheme to enable annotations with no arguments, as in @sparse above. @lsk567: Please have a look at those changes. If we decide that the 15% improvement is not worth the additional complexity, then we should cherry pick those changes.

The following PRs should be merged with this:

• C support for multiport iterator for sparse inputs reactor-c#95
• Python support for multiport iterator for sparse inputs lf-python-support#15

[lhstrh]

It left some nitpicks...