WIP: Readd shallow variables heuristic

[SSoelvsten]

Add the "shallow variables" heuristic anew to revert the slowdown introduced in #608 .

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.232%. Comparing base (fc67136) to head (c98eeb2).

Additional details and impacted files

@@              Coverage Diff              @@
##              main      #621       +/-   ##
=============================================
+ Coverage   97.225%   97.232%   +0.007%     
=============================================
  Files           90        90               
  Lines         6343      6358       +15     
=============================================
+ Hits          6167      6182       +15     
  Misses         176       176               

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[github-actions]

🟡 Regression Test (QBF 'breakthrough/3x4_19_bwnib')

'origin/internal/quantify/shallow_variables_heuristic' is a change in performance of -0.07% (stdev: 1.76%).

...	origin/main	origin/internal/quantify/shallow_variables_heuristic
Mean	24798.50	24815.25
Standard Deviation	254.11	437.71

Number of samples: 4

[github-actions]

🟢 Regression Test (Picotrav 'mem_ctrl')

'origin/internal/quantify/shallow_variables_heuristic' is a change in performance of 1.37% (stdev: 2.64%).

...	origin/main	origin/internal/quantify/shallow_variables_heuristic
Mean	116333.33	114740.67
Standard Deviation	3076.43	2836.94

Number of samples: 3

[github-actions]

🟢 Regression Test (QBF 'breakthrough/3x5_11_bwnib')

'origin/internal/quantify/shallow_variables_heuristic' is a change in performance of 1.22% (stdev: 1.63%).

...	origin/main	origin/internal/quantify/shallow_variables_heuristic
Mean	7505.40	7414.20
Standard Deviation	122.41	56.55

Number of samples: 5

[github-actions]

🟡 Regression Test (Picotrav 'arbiter')

'origin/internal/quantify/shallow_variables_heuristic' is a change in performance of -0.36% (stdev: 1.28%).

...	origin/main	origin/internal/quantify/shallow_variables_heuristic
Mean	62847.50	63075.25
Standard Deviation	807.07	343.97

Number of samples: 4

[github-actions]

🟡 Regression Test (QBF 'domineering/5x5_13_bwnib')

'origin/internal/quantify/shallow_variables_heuristic' is a change in performance of -2.44% (stdev: 4.65%).

...	origin/main	origin/internal/quantify/shallow_variables_heuristic
Mean	22471.17	23019.83
Standard Deviation	1046.03	120.64

Number of samples: 6

[github-actions]

🟡 Regression Test (Picotrav 'adder')

'origin/internal/quantify/shallow_variables_heuristic' is a change in performance of -0.03% (stdev: 0.63%).

...	origin/main	origin/internal/quantify/shallow_variables_heuristic
Mean	9166.40	9169.20
Standard Deviation	57.53	44.20

Number of samples: 5

[github-actions]

🟢 Regression Test (QBF 'ep/8x8_7_e-8-1_p-3-4_bwnib')

'origin/internal/quantify/shallow_variables_heuristic' is a change in performance of 59.51% (stdev: 8.60%).

...	origin/main	origin/internal/quantify/shallow_variables_heuristic
Mean	28429.60	11512.20
Standard Deviation	2444.69	81.77

Number of samples: 10

[SSoelvsten]

As can be seen from the above, this heuristic reverts the slowdown in #608 for ep/8x8. Now, the unit tests just need to be updated accordingly to make this mergeable.

[SSoelvsten]

Interestingly this does not seem to fully fix domineering/5x5_13_bwnib. So, we may also want to readd the heuristic based on the number of terminal arcs (or an improved version of it).