A Refactor of qcell to make it more generic, extensible, and checkable

[RustyYato]

see #11 for rationale

[RustyYato]

I have added ro, rw, rw2, and rw3 directly to the given owners and they just forward to the generic implementation on the trait. This is to preserve backwards compatibility. I will need to add in the other specialized functions to get complete back-compat, so don't merge just yet

[RustyYato]

TODO

• [ ] Move over documentation
  • [x] QCell
  • [ ] TCell
  • [ ] TLCell
  • [ ] LCell
• [ ] Move over tests
  • [x] QCell
  • [ ] TCell
  • [ ] TLCell
  • [ ] LCell
• [ ] Add documentation
  • [ ] QCell (new additions, unsafe constructors)
  • [ ] ValueCell
  • [ ] ValueCellOwner
• [ ] Complete back-compat
  • [ ] QCell
  • [ ] TCell
  • [ ] TLCell
  • [ ] LCell
• [ ] remove old code
• [ ] Add QCellOwnerID reclamation

Yay for long lists!

[RustyYato]

Most of the basic implementation work is done, I think I have addressed these concerns well enough. Is there anything that stands out to you @uazu?

[RustyYato]

Do you really need QCellOwnerID reclamation? I don't think there is ever a use case for creating 2^32 QCellOwners, and if you can think of one, we can just bump up to using u64 instead of u32. Then there won't be any use case of using 2^64 QCellOwners. (for scale, going and creating 1 million QCellOwners per second would take half a million years to exhaust u64)

If we ditch reclamation, we can support safe QCell on no_std

[RustyYato]

List of breaking changes:

• Removed QCellOwner::fast_new, this unsafe function no longer carries it's weight. QCellOwner::new will be just as fast in practice

[RustyYato]

Because we aren't going we this as per #11, close