Unimplemented simplified validation

[wlngai]

A full validation has been deployed on each benchmark run, which is quite heavy and verbose. Sometimes validation dataset might not be available. It will be easier to determine whether a benchmark execution has completed by checking if the output has the correct size.

[wlngai]

Before the validation step, the benchmark runner will first count the number of valid outputs. If the output size does not match expectation, then a completion failure (COM) will be thrown.