dependent PRs are not reported correctly

[eric-wieser]

When dependent PRs are made, then the heuristic of "look at all the changed files and attach them to this PR" doesn't work, because a file in a PR will also be attached to all its dependencies. This line in the make_port_status.py script then just picks the first one, essentially at random:

https://github.com/leanprover-community/mathlib4/blob/7e974fd3806866272e9f6d9e44fa04c210a21f87/scripts/make_port_status.py#L164

This also means that some PRs won't appear on the dashboard at all, because their association gets overwritten by a PR that depends on them.

cc @rwbarton, @jcommelin; reported by @mattrobball who noted leanprover-community/mathlib4#2332 didn't appear on the dashboard at all.

[eric-wieser]

The fix for this will likely need to live in the mathlib4 repo.

[mattrobball]

This PR is merged now so perhaps the behavior will change?

[mattrobball]

It now shows up as ported