Fix http 500 error during pdf upload (NullObject object has no attribute 'split')

[contributor]

This fixes http 500 error during upload of some books that pass pdf 1.4 validation at https://www.pdf-online.com/osa/validate.aspx, but cause the error in calbre-web. The error message is:

File "/app/calibre-web/cps/editbooks.py", line 1001, in edit_book_tags
    input_tags = tags.split(',')
AttributeError: 'NullObject' object has no attribute 'split'

Example of a book (21MB) that causes the error: https://github.com/contributor/calibre-web/blob/test-case/test/dotnet-aspire_cropped.pdf

NullObject comes from parsing metadata by pypdf/pypdf2/pypdf3 package.

This PR fixes http 500 error by ignoring 'NullObject /Keywords' and contains two commits:

1. the first commit does it “the right way”
2. the second commit does it “the simple way”: it avoids importing NullObject from three packages and localizes fix in single place

@OzzieIsaacs thank you for all your work! If you prefer commit 1, I will drop commit 2. If you prefer commit 2, you can just squash the PR during merge