On Forc.toml change, forc should try to update only package nodes that appear in the lock::Diff

[kayagokalp]

I will be simplifying and refactoring the code a little bit more so leaving this as a draft for now.

closes #900.

About the PR

Before this PR, forc was re-fetching all the dependencies once there is a mismatch between the lock file and the manifest.

For more context about the implementation please take a look into the discussion at the #900 and particularly this comment.

Basic Overview of What is Happening

1. Removed and added dependencies are detected
2. Removed dependencies handled by doing a BFS from the project node. Searching for detected removed dependencies and children of the removed dependencies (given that no other node depends on them) in the order of the breadth-first traversal.
3. Added dependencies handled by first connecting the added dependency to the root node and then recursively fetching_children for the dependencies of the newly added dependency while checking if any dependency of the newly added dependency is already fetched.

To test:

1. forc init --library child
2. forc init parent
3. forc build in the parent directory to build it before depending on the child.
4. add the "child" library as a dependency for the "parent"
5. forc build in the "parent" directory

This test procedure will be resulting in the following dependency graph:

[kayagokalp]

There are still some unnecessary calls to functions and duplicate code pieces, will be cleaning up and marking this as ready to review 🎉

[kayagokalp]

Realized a bug while removing the packages. The BuildPlan creation from lock fails and BuildPlan is always created from new once there is a package removal. Converting the PR into draft until that is solved.

[mitchmindtree]

Great work @kayagokalp! I've left a few comments but for the most part I think this is on the right track.

As a side-note: looks like there's a lot of duplicated code between BuildPlan::validate and BuildPlan::get_difference_with_manifest 🤔 Seeing as the only role of validate was originally to check if the package sets match, perhaps we can just rename validate to diff_manifest and alter it to return the difference in packages?

[kayagokalp]

@mitchmindtree I did the clean-up you asked for and I think currently it looks a lot better, thanks for the great review! Merging this will also close #1778, if the solution applied is acceptable.

[mitchmindtree]

Definitely much closer! I've added a few more things that should be addressed first

[mitchmindtree]

Looking great to me! Awesome work @kayagokalp :)

[JoshuaBatty]

Yeah this is really great @kayagokalp, nice work