Split fake::SyscallDriver implementations out of libtock_unittest

[jrvanwhy]

When I designed libtock_unittest, I intended for users of libtock-rs to be able to build their own fake::SyscallDriver implementations. Such implementations would be contained in crates with a dependency on libtock_unittest, and will therefore be unable to depend on pub(crate) items in libtock_unittest.

We currently have these drivers in libtock_unittest, which allows us to accidentally depend on pub(crate) items. I think we should move the fake::SyscallDriver implementations into their own crate, so our build system verifies that non-pub dependencies haven't cropped up.

I suggest the name libtock_fake_drivers for the new crate.

Do any of you have an opinion on this topic?

[hudson-ayers]

That sounds like a reasonable change to me! +1 to the name you suggested.