fix: correctly handle non-stale restarts

[piksel]

When restarting a container that had dependents, watchtower would previously treat those the same way as stale containers. This means that it would recreate them, even if the necessary information to do so was not present. It also means that it would attempt to clean up the image of the dependent, even though the same image would still be used. This may have been the cause of a number of issues experienced since v1.4.0 was released.

This PR adds checks that makes sure that linked containers also needs a valid config before being stopped, and that their images are only cleaned up if they are actually stale.

[codecov]

Codecov Report

Merging #1220 (b58b6f3) into main (70ca4c2) will decrease coverage by 0.19%.
The diff coverage is 63.63%.

@@            Coverage Diff             @@
##             main    #1220      +/-   ##
==========================================
- Coverage   62.70%   62.50%   -0.20%     
==========================================
  Files          23       23              
  Lines        1488     1507      +19     
==========================================
+ Hits          933      942       +9     
- Misses        471      475       +4     
- Partials       84       90       +6     

Impacted Files	Coverage Δ
internal/actions/update.go	63.07% <63.63%> (-6.03%)	⬇️
internal/flags/flags.go	88.82% <0.00%> (+0.17%)	⬆️
pkg/notifications/notifier.go	80.00% <0.00%> (+2.64%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 70ca4c2...b58b6f3. Read the comment docs.

[piksel]

There is one added safeguard that would require a lot of mocking, so I am letting it stay uncovered. Chances are that we don't even need the guard, that the problems it's protecting against was caused by the old cleanup-code.