Consider using an element index instead of word index for `dupw`

I wonder if it would make sense to refactor `dupw.n` so that `n` can be in range `0..=12` (as opposed to the current `0..=3`). It would then duplicate the word starting at the _element index_ `n` instead of the _word index_. I believe this would conceptually follow the recent refactor of `ProcessState::get_stack_word` that took a [word index in `0.17`](https://docs.rs/miden-processor/0.17.2/miden_processor/enum.ProcessState.html#method.get_stack_word) and takes an [element index](https://docs.rs/miden-processor/0.19.1/miden_processor/enum.ProcessState.html#method.get_stack_word_be) in the latest version.

The scenarios where this is useful is when you have a stack like this:

```
# => [num_inputs_start_ptr, num_inputs_end_ptr, INPUTS_COMMITMENT]
```

And you want to duplicate `INPUTS_COMMITMENT`. It's easy to do `dup.5 dup.5 dup.5 dup.5`, but `dupw.2` would just be more concise.

I think this would make `dupw` strictly more useful and all existing usages still work, by multiplying the previous `n` by 4.

But, it could be that I'm missing why this is not already allowed, so this is just a suggestion from a MASM user perspective.

Also, for consistency, I guess this should then also apply to other word instructions like `movupw` or `swapw`, but this then sounds like a much bigger refactor. So, if this is an unrealistic request, feel free to close the issue.