Make scoring code consistent with scoring procedure

[priyakasimbeg]

Currently the scoring code is not consistent with the scoring procedure.
The scoring procedure for a submission:

1. For each workload look at 5 trials per workload.
2. If >=3 trials have reached the target, use the best trial out of the 5 trials in terms of submission time.
3. If <3 trials have reached the target, the submission did not reach the target on that workload.

In scoring.py we have to make the following modifications:

1. Print a warning if fewer than 5 trials are found for a workload submission.
2. Check that at 3/5 trials have reached the target and only then use the best trial to score the submission (bug fix)
3. Print a warning fewer than 8 workloads are found.
4. Make sure the percentage workloads that reached targets is calculated over all 8 workloads as opposed to the number of workloads in the submission directory. Currently if the user scores a submission over an experiment with n workloads it will use n as the numerator.

[priyakasimbeg]

@runame do you have time to work on this?