Earlier the perf test runner using CPC was broken. This change fixes that.

With this change, we will be running CPC a little different than what we used to.
Earlier, we run CPC for 'n' of iterations and the first iteration should to be a
warmup iteration. Each iteration runs all the tests once.

Now, we run CPC for 'n' iterations for each test and 'n' is defined by each test.
The first iteration is, again, a warmup iteration and the rest are considered for the
metrics calculation. With this model, each iteration will run just one test instead of
running all the tests in an iteration. This helps in making the warmup iteration, a
true warmup iteration.

Minor changes:
Since the test csx returns an object now, the test related information has been
moved to the corresponding test csx files.