Improve documentation on non-independent sampling using greta_array()

It should be stated in the documentation of `greta_array()`, that using the `dim` argument with a scalar data argument results in identical copies of all elements and not independent ones. Compare the output of

```
set.seed(1)
x <- normal(0,1,dim=3)
simulate(model(x))

y <- greta_array(normal(0,1),dim=3)
simulate(model(y))
```

resulting in

```
$x
, , 1
[,1] [,2] [,3]
[1,] -1.036807 -0.414638 0.7169058

$y
, , 1
[,1] [,2] [,3]
[1,] 0.2973163 0.2973163 0.2973163
```