Divide and Multiply produce wrong results for Unsigned* types

As [discussed on gitter](https://gitter.im/imagej/imagej-ops?at=5a9683ca8f1c77ef3a664d18), the `math.divide` and `math.multiply` ops produce unexpected results when run on `Unsigned{Byte|Short}Type` data.

You can test with this Groovy script:
```groovy
#@ IOService io
#@ OpService ops

img = io.open("https://github.com/imagejan/sandbox/raw/wip/images/nucleus.png")
result = ops.run("math.divide", img, 32)
```

As @stelfrich [noticed](https://gitter.im/imagej/imagej-ops?at=5a9695af458cbde557daca0d):
> `Planar/ArrayImg` are backed by a `byte[]` and the op executes `plane[i] /= value`
for intensities > 127 that op will compute bogus results

The templates producing this code are here:

https://github.com/imagej/imagej-ops/blob/1375719d0ec1303601aef11d973ccbacda818580/src/main/templates/net/imagej/ops/math/ConstantToArrayImage.vm#L86-L91

https://github.com/imagej/imagej-ops/blob/1375719d0ec1303601aef11d973ccbacda818580/src/main/templates/net/imagej/ops/math/ConstantToPlanarImage.vm#L91-L96

In ImgLib2, the divide operation for `UnsignedByteType` is defined in their [`div()` method](https://github.com/imglib/imglib2/blob/af5f790c0fdd2b37bc8416dc1732ed57a6da145e/src/main/java/net/imglib2/type/numeric/integer/UnsignedByteType.java#L140-L143):

```java
public void div( final UnsignedByteType c )
{
set( get() / c.get() );
}
```

Maybe we should consider using the operators defined by the respective `Unsigned*` types, instead of doing calculations on the underlying `byte[]` arrays.

Start with the ConstantToArrayImage.vm and ConstantToPlanarImage.vm template sections linked in the issue, then reproduce math.divide with the provided Groovy script. Compare the generated operations with UnsignedByteType.div(); done means divide and multiply produce correct results for UnsignedByteType and UnsignedShortType values, including intensities above 127.

Written by the indexing model from the issue text.