Faster codesign_allocate

IIRC, codesign_allocate can take a while because it's moving so much data around, maybe because it has to make space at the start of the binary for an extra mach-o load command. We can probably bring this time down to < 10ms, by editing data at the start of the binary without having to make extra room