If the scanner detects new files with a path containing part of the name
of an already existing library item, the new item will incorrectly be
detected as being a parent directory of the already existing item and
the import will be aborted.
You can follow these steps to reproduce the issue:
```
❯ mkdir audiobooks/author/
❯ mv title\ 10 audiobooks/author
[2022-12-18 22:14:12] DEBUG: [Watcher] File Added /home/lars/dev/audiobookshelf/audiobooks/author/title 10/dictaphone.mp3
[2022-12-18 22:14:16] DEBUG: [DB] Library Items inserted 1
❯ mv title\ 1 audiobooks/author
[2022-12-18 22:15:03] DEBUG: [Watcher] File Added /home/lars/dev/audiobookshelf/audiobooks/author/title 1/dictaphone.mp3
[2022-12-18 22:15:07] WARN: [Scanner] Files were modified in a parent directory of a library item "title 10" - ignoring
```
Since `'title 10'.startsWith('title 1')` is `true`, the current code
makes this false assumption.
This patch fixes the issue by requiring a path separator to be part of
the matching path. This should ensure that only true parent directories
are detected.
This patch requires audiobookshelf to always use Unix file separators.
But that shouldn't be a problem since audiobookshelf always seems to use
these kinds of separators. Even on Windows.
The API for Google Books will return HTTP image URLs when matiching any
books using it as a search provider. In a secure environment, this
causes browser warnings.
All Google image links support HTTPS and we can safely switch to HTTOS
to avoid these warnings.
