The concept of an `Enumerator` replaces the existing concept of a
`BaseIterator`.  Furthermore, functionality is provided for enumerating
all combinations of a given set for a fixed value of `n`.

Property assertions are now operational again.  The key is that, for a
valid trace, all property assertions should hold.  However, for invalid
traces (i.e. where the given constraints fail) then we cannot expect a
property assertion to hold.

This puts in place a preliminary property-based testing framework which
will enumerate inputs and compare against a test oracle (as determined
by property assertions embedded within the constraints file).

This adds support for a "test generator" which automatically generates
test traces, and splits them using a given "oracle".  The latter is a
piece of Go code written specifically for this task.

This tweaks the test command so that it doesn't treat property
assertions as "absolute truth".  Rather, it attempts to find traces
which are accepted but where the property assertions don't hold.

At this stage, the system is fairly simplistic.  We need to think more
carefully about how to generate the traces.  This includes avoiding the
brute-force enumeration currently in play.  Likewise, thinking more
carefully about the "pool" of values used for testing.