Implement r0 crate in assembly

This implements the `r0::init_data` and `r0::zero_bss` routines in
assembly. There is a generic implementation for `riscv32` and
`riscv64`, since `riscv64` deals with alignment problems. The routines
are kept at their old calling site so that only one hardware thread
calls them. Consequently they are also inlined into the `start_rust`
function.

[Issue #122]