fix: asm() using the incorrect assembler for amd64 architecture

[CuB3y0nd]

Changed the priority of assembler usage in amd64

[Arusekk]

Thanks! Now I think I did not say it clearly: we want every architrcture to follow this pattern.

For example, assembling aarch64 should first try aarch64*-as, then native (if and only if host arch is aarch64), and then arm*-as as it frequently implements aarch64. And so on.

It should also be based on existing which_binutils function. My globs were supposed to be just illustrative, but the code should be kind of like the following:

aliases = {
 'aarch64': ['arm', 'something-exotic'],
 'amd64': ['i386', 'whatever'],
...
}

try(target)
if hostmachine == target: try(native)
for alias in aliases[target]:
  try(alias)

So just like it used to be, but native moved between the first try and the rest of them, instead of current buggy implementation (native is currently considered last).