First PR for pose estimation implementation (AUTH)

Created by: tsampazk

This is the first PR for the pose estimation implementation of the lightweight open pose algorithm.

The PRs for pose estimation would be split into three for easier reviewing due to the size of the implementation:

1. (This PR) Algorithm directory containing all relevant files of lightweight open pose which serve as a basis for the implementation
2. The learner implemented according to the project's specifications and a Target class required for the algorithms output
3. The demos directory that contains several use-cases of the implementation (this could be further split into several PRS). There are 7 demos that can be separated into 3 categories (3 Webots demos using TIAGo, 2 training and evaluation demos, 2 webcam test and webcam benchmarking demos)

Some notes about this PR:

1. I used the src dir as the root for the imports: e.g. from perception.pose_estimation.lightweight_open_pose.algorithm.modules.conv import *
2. I added the LICENSE file of the original repository that is Apache 2.0, please let me know if i need to separately add it inside the .py files, because they don't include it.
3. Two extra files were added by me that were not included in the original repo. These are with_mobile_net_v2.py and with_shufflenet.py. I modified existing implementations of the models to use them as backbones (i added the repo links in the files), so there is some work done there combining the original model implementations and lightweight open pose code to have the complete architecture. I am not really sure what license will be needed here, I added some license information in both, please let me know if it's ok.

Thank you!