Move linking into the offline trusted toolchain

[keithw]

If we had a Fix jump table mapped to a known set of virtual addresses, and Runnables are otherwise position-independent, we could pre-link all the Runnables and then directly mmap them read-only (treating them the same as any other Blob). Which would let us get rid of the in-memory cache of "linked" programs.

[Akshay-Srivatsan]

On inspection, the generated ELF files only have .rodata and .text segments, so they don't need any relocations for .bss or .data (which is necessary for the current linking approach to be correct, because those relocations would have to be per-run and not per-ELF-file). We could probably objcopy the code into flat .bin files, maybe with some minimal header at the beginning with a magic number and the offset of the entrypoint(s). The simplest table-based approach would be to pass the jump table address as part of the context, which should also be conducive to caching and branch prediction (modulo Spectre mitigations).