…ake a copy

of their bytelists eliminates test failures.  I had considered doing this
anyways since it will not carry on an entire source line with it but at the
same time I hoped to not do the copy to save some up front work (e.g. creating
a new array and copying into it).

This is the problem I observed running spec:compiler.  When run as a full suite
then I see some failing tests.  If I look at one failing one I can see source
like:

```ruby
a.foo = 1
```

becomign IR like:

```text
  04:     %v_4 := call_0o(*a, callType: NORMAL, name: foo, potentiallyRefined: false)
  05:     %v_5 := call_1f(%v_4, fix<1>, fixNum: 1, callType: NORMAL, name: ==, potentiallyRefined: false)
```

Notice that is calling 'foo' and not 'foo='????  If I extract this spec into its
own file then I see:

```text
  05:             attr_assign(*a, fix<1>, callType: NORMAL, name: foo=, potentiallyRefined: false)
```
WTF is going on here?  how would not copying and sharing the same backing line
end up causing this but only if run within a larger file?  The best I can figure
is another spec is doing something destructive to the backing array without
realizing it is shared.