It appears to be looking for and capturing lines containing page numbers in one of 4 forms. Roughly:

'      nnn   <page>   '
'    - nnn - <page>   '
'      A-nnn <page>   '
'  page nnn of mmm    '
[download]

This becomes fairly clear if you break up and expand the regex a little using /x.

my $pattern = qr[
    (
            \n\s*   [0-9]{1,3}          \s*\n\s* <page>       \s*\n
        |    
            \n\s* - [0-9]{1,3} -        \s*\n\s* <page>       \s*\n
        |
            \n\s*   [A-Za-z]-[0-9]{1,3} \s*\n\s* <page>       \s*\n
        |
            \n\s*  page \s* [0-9]{1,3}  \s* of \s* [0-9]{1,3}  \s*\n
    )
]x;
[download]

There are many ways the regex could be improved, but that isn't what you asked :)