Add script to calculate specializations using dynamic programming

[LorenzSchueler]

This PR adds a Python script to compute the best node specializations based on node counts obtained from the statistics module using a dynamic programming approach.
It produces the same results as the ILP implementations, but for num_specializations >= 4 it is significantly faster.

I did not incorporate this into the existing script, because the used data types are different. This can be done in another PR if needed.

The output looks like this:

--- Node type: Inner ---
Num Specializations: 1
Min cost: 214650615200
Specializations: [256]

Num Specializations: 2
Min cost: 154199374790
Specializations: [102, 256]

Num Specializations: 3
Min cost: 141878721821
Specializations: [9, 105, 256]

Num Specializations: 4
Min cost: 135478558857
Specializations: [4, 78, 109, 256]

Num Specializations: 5
Min cost: 132960791105
Specializations: [4, 68, 91, 113, 256]

--- Node type: Leaf ---
Num Specializations: 1
Min cost: 503024231374.0
Specializations: [256]

Num Specializations: 2
Min cost: 74731897003.0
Specializations: [7, 256]

Num Specializations: 3
Min cost: 55646058849.0
Specializations: [4, 149, 256]

Num Specializations: 4
Min cost: 46106649717.0
Specializations: [2, 26, 149, 256]

Num Specializations: 5
Min cost: 43565650674.0
Specializations: [2, 16, 91, 149, 256]

This is the output when DEBUG and PRINT_TABLE are set to True

Num Specializations: 1
Min cost: 850
Specializations: [4]

10:[0]    40:[1]    210:[2]   360:[3]   850:[4]   
          INF:[]    INF:[]    INF:[]    INF:[]    
                    INF:[]    INF:[]    INF:[]    
                              INF:[]    INF:[]    
                                        INF:[]    
Num Specializations: 2
Min cost: 710
Specializations: [2, 4]

10:[0], INF:[]       40:[1], INF:[]       210:[2], INF:[]      360:[3], INF:[]      850:[4], INF:[]      
                     INF:[], 30:[0, 1]    INF:[], 190:[0, 2]   INF:[], 330:[0, 3]   INF:[], 810:[0, 4]   
                                          INF:[], 190:[1, 2]   INF:[], 320:[1, 3]   INF:[], 790:[1, 4]   
                                                               INF:[], 290:[2, 3]   INF:[], 710:[2, 4]   
                                                                                    INF:[], 760:[3, 4]   
Num Specializations: 3
Min cost: 690
Specializations: [0, 2, 4]

10:[0], INF:[], INF:[]              40:[1], INF:[], INF:[]              210:[2], INF:[], INF:[]             360:[3], INF:[], INF:[]             850:[4], INF:[], INF:[]             
                                    INF:[], 30:[0, 1], INF:[]           INF:[], 190:[0, 2], INF:[]          INF:[], 330:[0, 3], INF:[]          INF:[], 810:[0, 4], INF:[]          
                                                                        INF:[], 190:[1, 2], 180:[0, 1, 2]   INF:[], 320:[1, 3], 310:[0, 1, 3]   INF:[], 790:[1, 4], 780:[0, 1, 4]   
                                                                                                            INF:[], 290:[2, 3], 270:[0, 2, 3]   INF:[], 710:[2, 4], 690:[0, 2, 4]   
                                                                                                                                                INF:[], 760:[3, 4], 690:[2, 3, 4]
                                                                                                                                                                                    ```

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
see 3 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[b-scholz]

LGTM