Define some test cases for library scanning

[hatzel]

We are in the process of building an integration test suite library migrations (testing the process of a user changing files between scans of the filesystem).
We will just define some test cases in natural language in this issue and check them off as the test is implemented.

• Recover from deletion: the first state has a book, second doesn't, it's back in the third state
• Book get's renamed
• Work with moved audiobooks:
  • In the first state the book is called 'lol', in the second it is called 'rofl', file contents are the same
  • In the first state the book is called 'lol', in the second it is called 'rofl', file contents are the same but there is a new book also called 'lol'
    • The hash should probably get precedence
• Book with same name has different file contents
• Book get's deleted, shows up with different content again
  • What should even happen here if we see a book with the same name but different hash? I think that when in doubt we should keep the playstates. Manually starting playback from 0 isn't exactly hard to do, but finding the time code you last listened at could be really annoying.
• Multifile book is updated.
  • Make sure it's remuxed
• Multifile audiobook gets new file
• Make sure it's remuxed again