[FEATURE] Alternative approach to page allocation

[jorisdral]

#301 implemented an alternative approach to our custom disk page allocator. The PR got stale and never landed on the main branch, but it would probably still be useful to change to the alternative implementation.

Quote from @dcoutts about the alternative approach:

I still think it was a better idea (simpler and faster), but indeed it would need other changes to allow multi-page values to use non-contigious 4k pages.