setnafill error for invalid input should be more helpful

Related: #4113, #3992 

I find this a bit confusing:

```r
setnafill(iris, fill=0)
# Error in setnafill(iris, fill = 0): 'x' argument must be numeric type, or list/data.table of numeric types
```

It would be better to be more specific -- it failed because the input has some non-`numeric` columns (and possibly name the problem columns, or at least one of them).