feat(ext/ffi): Make op_ffi_ptr_of fast

[aapoalas]

Makes op_ffi_ptr_of fast. One of the tests changed from printing false to true as the fast &[u8] slice path creates the slice with a null pointer. Thus the op_ffi_ptr_of will now return a null pointer value whereas previously it returned a dangling pointer value.

[littledivy]

#[op(fast)]
fn op_ffi_ptr_of<FP>(
  state: &mut deno_core::OpState,
  buf: &mut [u8],
  out: &mut [u32],
) -> Result<(), AnyError>
where
  FP: FfiPermissions + 'static,
{
  check_unstable(state, "Deno.UnsafePointer#of");
  let permissions = state.borrow_mut::<FP>();
  permissions.check(None)?;

  let ptr = buf.as_ptr() as u64;

  out[0] = ptr as u32;
  out[1] = (ptr >> 32) as u32;
  Ok(())
}

out[0]/out[1] will panic out of bounds so the length check is also not needed.

[aapoalas]

Is it preferable that a panic occurs or should it rather be a throw?

[aapoalas]

Also, I would presume that it'd be faster to write in a single "operation" rather than do the shifting etc. I was actually thinking of perhaps making the checks debug_asserts.

[littledivy]

yes panic is better when its not possible to invoke this from a public API. It forces users to report bugs

[aapoalas]

I changed these to asserts and went with a write-to-pointer-of-usize handling instead.