Projects dependency map

[wmalpica]

We have a lot of big projects that we are currently working on. These are mostly all going to be work on based off of the UCX branch.
This work will be based off of the UCX brand because it already contain a lot of changes all over the code base which you want to keep.

There a big three key projects and each of these have certain common requirements. In order of priority they are:
Communication c++ layer: #926
Batch Job Executor: #807
Avoid recalculating nodes in a graph: #696

The communication project needs to be completed first in order for us to be able to merge in the UCX branch into our main development branch.
The batch job executor project is also key because it will enable us to really scale and manage our memory for TPCX-BB.
These two projects should be tackled in parallel while properly managing their dependencies.

The batch executor will require a lot of refactoring of a code base and reorganizing how the kernels execute their work.
In order to make some of the refactoring work easier "Barriers Required for Distributed execution" #923 should be done first.

Additionally the code and of all the kernels should be retractored such that you can execute all the work using a run_batch function. All this refectoring could then be tested and validated to make sure it doesn't break anything. After the code of the kernels has been refactored then we can actually create a batch executor. Part of the code refactoring may be small projects in of themselves. For example calculating a partition plan for order by should be refractored into its own kernel. The join kernel will also require some non-trivial refactoring.

In the first implementation of the batch executor we can continue to use the same cache data the same ownership paradigm that we have with blazingTable. Cache data and blazing table can change afterwards when we make changes to avoid recalculating nodes in a graph. The new data ownership paradigm is defined in this project:
#924