flakes: failing follows test case

[edrex]

this is a bug report in the form of a failing test case

A has input B, B has input foobar, and

• B: inputs.D.follows = "foobar/D"
• A: inputs.B.inputs.foobar.follows = "foobar"

Result:

error: input 'B/D' follows a non-existent input 'B/foobar/D'

Expected: ['foobar', 'D']

Tested with v2.6.1 with #6036 cherry-picked, since master is giving me a SIGBUS right now.

[edrex]

cc @balsoft

[edrex]

Context: found when attempting to use https://github.com/srid/emanote/blob/master/flake.nix as an input. @srid

[edrex]

I made a small change to the existing test case to trigger the failure, but it seems like factoring out the flake follows tests to be more explicit should make it easier to reason about expected behavior.

Very similar to #6013 but not quite the same.

Maybe a common cause with #5790