Performance improvement when finding tasks in a subworkflow

[jbirddog]

Before a subworkflow would use an exception based approach to finding a task by id. When profiling an example multi-instance workflow this could result in ~1 million exceptions being thrown just so None could be returned. Fix is to just directly call get on self.tasks. For the example process this improves runtime by about 22% and results in ~3 million less function calls. Timings below are from spiff-arena's backend.

Before:

         19118665 function calls (18030458 primitive calls) in 56.559 seconds

   Ordered by: call count

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
  1801934    2.268    0.000    2.268    0.000 {built-in method builtins.hash}
  1798819    4.427    0.000    6.690    0.000 uuid.py:268(__hash__)
1749267/1749091    3.280    0.000    5.986    0.000 {method 'get' of 'dict' objects}
  1344447    1.769    0.000    1.780    0.000 {built-in method builtins.isinstance}
  1047694    2.238    0.000    2.346    0.000 uuid.py:280(__str__)
  1047398    6.325    0.000   18.235    0.000 workflow.py:142(get_task_from_id)
  1037940    3.476    0.000   22.200    0.000 subworkflow.py:64(get_task_from_id)
  1027950    3.296    0.000    5.244    0.000 exceptions.py:42(__init__)
  1027950    1.939    0.000    1.939    0.000 exceptions.py:24(__init__)

After:

         15938976 function calls (14850242 primitive calls) in 44.234 seconds

   Ordered by: call count

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
2773162/2771613    5.985    0.000   12.557    0.000 {method 'get' of 'dict' objects}
  1791937    2.241    0.000    2.241    0.000 {built-in method builtins.hash}
  1788829    4.373    0.000    6.677    0.000 uuid.py:268(__hash__)
1341144/1340474    1.770    0.000    1.780    0.000 {built-in method builtins.isinstance}
  1037940    2.586    0.000    9.273    0.000 subworkflow.py:64(get_task_from_id)
   423337    1.129    0.000    1.806    0.000 registry.py:51(clean)
419712/8329    2.792    0.000    8.322    0.001 dictionary.py:87(convert)
419712/8329    1.816    0.000    8.464    0.001 registry.py:39(convert)
   293445    0.370    0.000    0.370    0.000 task.py:139(has_state)

[essweine]

I have concerns about this PR so do not merge yet.

[essweine]

There was originally a reason for this (I think it had to do with error handling) but I don't think it's relevant at this point.