Cranelift: pass iterators to ABIMachineSpec::compute_arg_locs (#5095)

Instead of slices. This gives us more flexibility to pass custom sequences
without needing to allocate a `Vec` to hold them and pass in as a slice.

cranelift/codegen/src/isa/aarch64/abi.rs

@@ -87,13 +87,16 @@ impl ABIMachineSpec for AArch64MachineDeps {
         16
     }
 
-    fn compute_arg_locs(
+    fn compute_arg_locs<'a, I>(
         call_conv: isa::CallConv,
         _flags: &settings::Flags,
-        params: &[ir::AbiParam],
+        params: I,
         args_or_rets: ArgsOrRets,
         add_ret_area_ptr: bool,
-    ) -> CodegenResult<(ABIArgVec, i64, Option<usize>)> {
+    ) -> CodegenResult<(ABIArgVec, i64, Option<usize>)>
+    where
+        I: IntoIterator<Item = &'a ir::AbiParam>,
+    {
         let is_apple_cc = call_conv.extends_apple_aarch64();
 
         // See AArch64 ABI (https://github.com/ARM-software/abi-aa/blob/2021Q1/aapcs64/aapcs64.rst#64parameter-passing), sections 6.4.