Use memmap for fs cache layer

[ChillFish8]

Currently, lnx-fs uses a in memory cache which uses a least frequently used policy, this has some good advantages but currently we have two large issues:

• Memory is not aligned
• We cannot incrementally build up a file in the cache as a single contiguous (to rust at least) slice.

The goal of using memmap here is to allocate using anonymous memory the entire file, but only write and page in sections we currently have "cached".

This means the direct IO workers are still doing the reading and writing (which improves our support for NFS and similar block devices) while allowing use to fix the aforementioned issues.