replace values in a variable greta array?

This errors:
```r
x <- normal(0, 1, dim = 3)
x[1] <- 0
```
> ```
> Error: cannot replace values in a variable greta array
> ```

But it needn't, we're just creating a new greta array `x` which is an operation.

I'm not sure whether allowing this would make for a clearer or a less clear interface. The first random variable will still be floating around, and its density will be included in the joint model (though as an unconnected independent variable).

This issue is just a reminder to think about this in the future. Any input very welcome!