fix(GenerateDAG): new algorithm

[antoinegelloz]

• Fix the algorithm assigning files to the docker build context of each nodes of the graph
• Reduce cyclomatic complexities of the GenerateDAG function and its sub-functions

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 79.45205% with 30 lines in your changes missing coverage. Please review.

Project coverage is 59.44%. Comparing base (dde85cd) to head (b109198).
Report is 30 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/dib/generate_dag.go	79.45%	17 Missing and 13 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #556      +/-   ##
==========================================
+ Coverage   58.80%   59.44%   +0.63%     
==========================================
  Files          51       51              
  Lines        2350     2387      +37     
==========================================
+ Hits         1382     1419      +37     
+ Misses        861      860       -1     
- Partials      107      108       +1     

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

[joulaud]

I see the algorithm was completely rewritten from scratch and thus the diff is not meaningfull.

To ease review, could you outline :

• the bug you wanted to fix
• the general working of the new algorithm and some quick justification for choosing it
• perhaps some comparison with the old one (stating why the old one was beyond repair and, if there is some, the changes in behaviour and results)
  ?

Even best if the meaningful bits are in commit message to be remembered eternally in Git history.

Thank you for your hard work.

[antoinegelloz]

I see the algorithm was completely rewritten from scratch and thus the diff is not meaningfull.

To ease review, could you outline :

• the bug you wanted to fix
• the general working of the new algorithm and some quick justification for choosing it
• perhaps some comparison with the old one (stating why the old one was beyond repair and, if there is some, the changes in behaviour and results)
  ?

Even best if the meaningful bits are in commit message to be remembered eternally in Git history.

Thank you for your hard work.

I understand your concern, thank you.
I'm closing this one because we cannot review it for now. I just created this new PR, which is improving the testing of the current algorithm, to show the different problems with it.