Replace toposort with graphlib

[SajidAlamQB]

Description

Related to: #1820

This ticket aims at dropping toposort dependency in favor of built-in graphlib if possible.

Development notes

QA notes

Checklist

• Read the contributing guidelines
• Opened this PR as a 'Draft Pull Request' if it is work-in-progress
• Updated the documentation to reflect the code changes
• Added new entries to the RELEASE.md file
• Added tests to cover my changes

[rashidakanchwala]

LGTM, and the tests are all passing.

So Toposort performs several tasks under the hood, such as determining the layer and parent layer relationships. But, with graphlib, we must explicitly ensure that we don't add a layer as its own parent. Interesting.

[ravi-kumar-pilla]

Hi @SajidAlamQB ,

Nicely done. I was testing the PR and I see there is a minor issue in the UI. The order of layers is shifted -

Layers feature and model_input are shifted. Even the DAG has Feature Importance Output dataset after Model Input Table. Can you please test it on your side. Thank you

Before

After

[SajidAlamQB]

Layers feature and model_input are shifted. Even the DAG has Feature Importance Output dataset after Model Input Table. Can you please test it on your side. Thank you

Hey @ravi-kumar-pilla, this should be fixed now let me know if its all fine.

[ravi-kumar-pilla]

Hey @ravi-kumar-pilla, this should be fixed now let me know if its all fine.

Hi @SajidAlamQB , I tested it now and the order looks good. Thank you for that. May I know why was the static order different or why - return sorted(sorted_layers, key=lambda layer: (sorted_layers.index(layer), layer)) was needed.

[ravi-kumar-pilla]

LGTM ! It would be nice to have this in the release note to track the changes. Thank you @SajidAlamQB

[SajidAlamQB]

Hey @ravi-kumar-pilla, this should be fixed now let me know if its all fine.

Hi @SajidAlamQB , I tested it now and the order looks good. Thank you for that. May I know why was the static order different or why - return sorted(sorted_layers, key=lambda layer: (sorted_layers.index(layer), layer)) was needed.

The order of layers by the TopologicalSorter was decided by their insertion order. Since the insertion order was not stable it led to unexpected ordering in the output. So adding that final sort helped maintain the original input order. I will add some comments here for future reference as well.

[astrojuanlu]

LGTM! 👏🏼