Use __dependents_rules__ to optimize pants dependents calculation

[grihabor]

Is your feature request related to a problem? Please describe.
We have the following monorepo structure:

src/python/ # shared code
project1/
project2/
...

Projects are not allowed to depend on different project targets except for src/python, so we have rules for every project:

__dependents_rules__(
    # We don't want anybody to import from project1.
    (
        {"path": "project1/**"},
        "project1/**",
        "!*",
    ),
    ("*", "!*"),
)

Every project has it's own docker_image targets and helm_deployment targets.
Now, imagine we change some code in project1/src/ and try to deploy it with:

pants dependents --transitive --changed-since=origin/main

This triggers a bunch of calculation like:

⠁ 0.00s Rendering Helm deployment project1/helm:project1
⠁ 1.03s Rendering Helm deployment project2/helm:project2
⠁ 1.45s Rendering Helm deployment project3/helm:project3
⠁ 1.52s Rendering Helm deployment project4/helm:project4
⠁ 0.54s Rendering Helm deployment project5/helm:project5
⠁ 1.61s Rendering Helm deployment project6/helm:project6
⠁ 0.09s Rendering Helm deployment project7/helm:project7

which takes about 4 minutes without cache.

However we know for sure that no project can depend on project1/src/ which means we can skip other projects.

Describe the solution you'd like
Collect information from __dependents_rules__ and use it to optimize pants dependents logic.

Describe alternatives you've considered
Live with very slow pants dependents and cry