Possible optimisations in UVM Makefile

[rswarbrick]

This issue separately tracks a specific part of issue #674.

The Makefile in dv/uvm/core_ibex/Makefile tries to be careful to regenerate targets if you change an option. This way, we ensure that if you run make twice with the same test (say) but change options, the test will actually be rebuilt and re-run.

One unfortunate side-effect is that the second command in

make SEED=123 TEST=foo
make SEED=123 TEST=bar

will start by deleting the results for the foo test. (As far as the Makefile is concerned, TEST is just another option that affects the output of the test generation stage).

This issue is to track whether we can avoid that somehow. Also, maybe there are other rebuilds we can avoid, such as rebuilding the instruction generator when you've just changed the test list.