implement in-memory volatile filesystem

[XiaochenCui]

We want an "in-memory volatile filesystem" that:

• Demonstrates how to implement a custom "filesystem provider".
• Store temporary data that should be automatically discarded on server reboot. (like the /tmp dir on Linux)
• It supposed to be much faster than the default "puter fs provider". A benchmark is desired for assessment.

TODO

• Implement all filesystem APIs.
  • quick_check
  • stat
  • readdir
  • move
  • copy_tree
  • unlink
  • rmdir
  • mkdir
  • write_new
  • write_overwrite
• pass apitest
  • write_cart
  • move_cart
  • copy_cart
  • write_and_read
  • move
  • stat
  • readdir
  • mkdir
  • batch
  • delete
  • telem_write
• ACL
  • Remove hardcoded permission checks, use standard ACL.
  • Make it mountable at arbitrary location (e.g., /tmp, /dev/shm). Currently, it can only be mounted under /<username>/tmp to run the tests.
• Store file path relative to the mountpoint, so remounting to a different point can be done easily.
• Make it works on multi-server deployment. (We won’t be working on this since the main purpose of the "memory FS provider" is for demonstration, and it should remain as simple as possible.)
• test clutter
  • make the test result more readable
  • run "nightly-build" and push the test result to slack
  • A benchmark to assess if it's much faster than "puter fs provider"
• in the future
  • write sophisticate docs for it, and tutorial for "write your own fs provider"
  • move implementation details to 1 files, and avoid injection to other files

[KernelDeimos]

The remaining filesystem methods and features can be included in this PR. Let me know when new methods are added and I'll test on this PR's branch directly.