refactor the search for direct-origin dependencies

[dimbleby]

refactoring suggested by #5903.

• all the search_for_xxx() methods returned a list of length one, which is a bit weird
• the whole pattern is repeated three times, let's just do it once.

IMO it's kinda ugly that search_for_xxx(dependency) makes updates to dependency, that doesn't match my intuition for what such a method ought to do. But trying to refactor that broke more tests than I care to investigate.

In particular this rearrangement ensures that the updates that we make to the input dependency are the same regardless of whether we got there via the _deferred_cache or not.

[radoering]

Have you thought about the following changes in behavior? (Not sure about the consequences or if it's relevant at all.)

package is not cloned anymore for file and directory dependencies. Thus, package.root_dir, package.files, package.develop are set according to the first dependency.

It seems like VCS dependencies were only put into the cache but never taken from it. Seems like a bug, but had the same effect as cloning. Now, package.develop is set according to the first dependency.

[dimbleby]

@radoering yes I noticed those things, but couldn't think of a reason that they should matter - so chose to trust the test suite when it said that it saw no problem.

[radoering]

Me neither. Should be fine.

[github-actions]

This pull request has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.