minor infidelity when lifting clz

[regehr]

so LLVM's cttz and ctlz intrinsics have a flag to indicate whether the behavior should be undefined when the main argument is zero.
my understanding is that arm64's clz is defined in the case of zero argument, meaning that clz should be lifted to ctlz x, false
however, anvill/remill lift clz to ctlz x, true

in other words, this function:

0000000000000000 <slice>:
   0:	dac00008 	rbit	x8, x0
   4:	dac01100 	clz	x0, x8
   8:	d65f03c0 	ret

expressed like this:

{
  "os": "linux", 
  "functions": [
    {
      "return_stack_pointer": {"register": "SP", "type": "L", "offset": 0}, 
      "name": "slice", 
      "parameters": [
        {"register": "X0", "type": "l"}
      ], 
      "return_address": {"register": "X0", "type": "L"}, 
      "return_values": [
        {"register": "X0", "type": "l"}
      ], 
      "address": 0
    }
  ],
  "arch": "aarch64",
  "stack": {"size": 24576, "start_offset": 4096, "address": 87960930222080},
  "memory": [
    {
      "is_writeable": false, 
      "data": "0800c0da0011c0dac0035fd6",
      "is_executable": true,
      "is_readable": true,
      "address": 0
    }
  ]
}

decompiles to this:

define i64 @slice(i64) local_unnamed_addr #0 {
  %2 = tail call i64 @llvm.bitreverse.i64(i64 %0) #3
  %3 = tail call i64 @llvm.ctlz.i64(i64 %2, i1 true) #3, !range !0
  %4 = tail call %struct.Memory* @__remill_function_return(%struct.State* nonnull undef, i64 0, %struct.Memory* null) #3
  ret i64 %3
}

the ARM code is defined for argument zero whereas the LLVM code evaluates to poison in this case.

the fix is trivial: when lifting this ARM instruction, set the UB flag to false