Improve qualification tool estimation

[nartal1]

While working on issue #385 , I came across few issues which if addressed would make the qualification tool speedup estimation better.

1. Some Execs do not have metrics populated, so while calculating the number of Execs within a stage we miss these. We map stageToExec based on accumulatorId and since metrics are not populated, we miss adding those Execs to that corresponding stage - [FEA] Qualification tool - enhance exec to stage mapping  #615

2. We don't consider duration of each Exec within a stage i.e we evenly distribute durations(total taskMetrics durations) to all the known Execs. The issue is one Exec with speedup 5x maybe taking 90% of the total time and other Exec with 2x could be taking the remaining 10%. But in our case we assign same durations to both and take the average of the speedups. I am not certain if there is a way to fix this but would be good to investigate.

3. promote_precision is shown as Not supported expression. Should fix this as it is supported

[amahussein]

@nartal1 I moved the promote_precision to a separate issue #517 .

[nartal1]

1. Fixed by - PR-634

2. I investigated this further and we cannot accomplish this due to the way metrics are generated per stage. Typically, CPU operators are consolidated within a single stage without detailed breakdowns of individual durations. This grouping constrains the granularity of timing details, as it does not precisely measure the time taken by each specific operator. So the current implementation which is to assign equal durations to all the operators in a given stage seems to be the best possible solution in this case.
   Closing this issue