Transpose2d, Concatenate2d, and up to 3 Clones for io_stream

[jmduarte]

This PR has 3 independent, minor changes (I can split them into smaller PRs if needed):

• Transpose2d for io_stream
• Concatenate2d for io_stream
• Clone streams up to 3 times (instead of just up to 2 times)
• Add unit test for all 3

These features are needed, for example in this interaction network model: https://gist.github.com/jmduarte/9ea95df65feac29f9663c47a439e2d7e

[thesps]

Thanks for this, looks nice! To be clear, for the review, only the increase from 1 to 100 test examples is a requested change, the other items are for discussion.

I tested with the 100 examples and it still works fine. I ran cosim as well since the io_stream version could have some dataflow behaviour: all good there too.

Maybe something for the future: now we have layers to clone a stream 2 times and another to clone it 3 times. Can there be a CloneN solution?

[jmduarte]

@thesps I responded to your comments and added some follow-up questions.

We should probably have a CloneN solution, but it's not clear to me how to handle the HLS templates for that.

[jmduarte]

@thesps @vloncar let me know if there are any other updates needed or if you have an idea of how to do the CloneN solution

[jmduarte]

Added concatenate1d here to partially resolve #436

[jmduarte]

@thesps @vloncar rebased this PR and I think it's much cleaner now and good to go!

Raised issues have been resolved