Review boot memory mapping.

[kubanrob]

While working on the ihk support, @gypsephi found some mysterious code.

For example, when initializing the kernel space, a page is set invalid using a magic number. Its the first page after the 8 pages already set invalid. But why 8 pages. At least a comment is missing here.

Similarly, it should be documented why we have two pml4 page tables. (The second one is the one we use in the kernel, the first one is used be the bootstrap/start code).

I suggest we review/document this in parallel to the cleanup of the ihk branch. We might use this issue to collect questions and answers related to this.