generic page pool for MMU-based systems

[andrewboie]

Is your enhancement proposal related to a problem? Please describe.
We have some use-cases for a page pool:

1. Pages to allocate MMU data structures
2. Pages for anonymous memory mappings
3. Pages to increase the size of the system heap arena with sbrk() which is really just a special case of This is a test commit #2

Right now on x86 we have:

• a separate pool for MMU paging structure pages which has to be tuned in a cumbersome way; the number you need depends on how many active memory domains are required and how much memory has been mapped besides system RAM.
• For the newlib libc, we do not memory map the heap. We have a fixed sized physical arena from the _end symbol to the bounds of RAM. sbrk() calls just increment a pointer within this area. No attempt is made not to crash into data structures populated by firmware and we are not using available memory in the first megabyte.
• No support for anonymous memory mappings yet

Describe the solution you'd like
A page pool to draw page frames from for all the above use-cases.
All these pages are either getting memory mapped into the virtual address space, or are used for paging structures, there is no need for contiguous physical regions larger than a page.
Will need to play nice with demand paging, with ability to anonymously map more memory even when all RAM frames are in use.

[dcpleung]

Already done. (commit b011771)