Add ICAROS Python Bindings

[sniyaz]

This PR adds in the Python bindings from https://github.com/icaros-usc/libada. The ultimate goal is to replace their stack with ours, while still conforming to the same Python API. Since most of ICAROS's work goes through the Python bindings (as far as I know) it shouldn't matter that PRL's C++ guts are much more up-to-date.

This PR essentially 1) adds in the binding source files and then 2) fixes everything to build (switching over methods that we've deprecated in AIKIDO, finding pybind11 in a more-general way, adding methods to the Ada.hpp interface that ICAROS added, etc). I also think I've fixed a few bugs in the original bindings (or at least what I perceived as bugs 😅).

There are also two large outstanding issues:

1. The version of pybind11 used in these bindings (and it looks like the version used in our aikidopy bindings as well) is 2.2.0. Most of PRL currently uses Ubuntu 18.04, which packages pybind11 2.0.1. Downgrading the pybind11 version is messy and doesn't seem to make sense: my gut is to 1) make building the bindings optional and 2) provide clear documentation on how to install a proper version of pybind11 from source (it's actually incredibly easy).

2. More critical: right now it looks like ICAROS is using ROS1, whereas PRL one day wants to move to ROS2. Unfortunately, because ROS1 doesn't support Python 3 I've had to set the Python version at 2.7. There are other barriers to moving to Python 3: installing python3-rospkg seems to uninstall all of the ROS melodic packages, which is definitely a big deal. I know @egordon was really set on setting Python 3, so this is something we need to debate. It seems like 2.7 is ugly, while 3.6 is potentially a show-stopper for ICAROS.

Testing Plan

I ran simple_trajectories.py, and it works great! :)

---

Before creating a pull request

• Document new methods and classes
• Format code with make format

Before merging a pull request

• Add unit test(s) for this change

[hejia-zhang]

Looks good!

[hejia-zhang]

1. We could probably rename ComputeJointSpacePath as ComputeRetimedJointSpacePath? 2. We could change the compute_retime_path function to only accept aikido::trajectory::Interpolated typed trajectory?
   I have tested this branch with another lab assignment code and it works great!

[brianhou]

I mostly have some quick high-level questions:

• Would it make sense to move src/adapy to a top-level adapy directory or python/adapy? Otherwise, the layout of src and include no longer match our other repos. (python/adapy would match python/aikidopy; it might also then make sense to move the Python scripts into, say, python/tests.)
• Should we organize this PR into two? One with actual libada changes, one with bindings.
• Do bindings need to be defined as C++ lambda expressions? I feel like this would be way more readable if, say, we explicitly named the create_ik function and then did all the module m.def stuff with those named functions.

[brianhou]

Do bindings need to be defined as C++ lambda expressions? I feel like this would be way more readable if, say, we explicitly named the create_ik function and then did all the module m.def stuff with those named functions.

I'm willing to accept the changes as is for now, but I still feel like this would be better.

[sniyaz]

An update: I've (finally) managed to get my dev box set up with 20.04 + Noetic. I can confirm that from a fresh, standard machine install everything works fine. The pybind11 version is more than sufficient and everything works great in sim :)

I'm going to work on some more of this feedback from @brianhou now that I have the new env set up and working 😃

[sniyaz]

@brianhou and others, I've responded to the lambda feedback and cleaned up a few other things. I think we're ready for a final review?

I wasn't sure whether to use camel case for the named methods since that seemed a little confusing, but happy to change it if we want :)

[hejia-zhang]

It looks good!

[egordon]

Everything still runs perfectly on the real bot. Merging as admin past the Travis checks. Good job all!