Improved Coverage Testing

[ncough]

A number of bugs have slipped past the coverage tester as we start to worry about the tail end of architecture behaviours. It would be nice if the coverage tester could be extended, such that:

• It's easier to configure the initial state for differential evaluation, e.g., perhaps by running nominated ASL function that sets the state as necessary.
• Alternatively, it would be nice to know if the evaluation is interacting with state that has not been meaningfully initialised.
• It's possible to have memory loads return meaningful values, e.g., by overriding the memory load call to return random values during differential testing.
• The evaluator can collect a trace of function calls, that can then be compared between the two evaluations.
• It's possible to nominate two sets of overrides, one for differential testing and another for partial evaluation.