Use a common code base for both CLI and Colab notebook implementations

[RomainAbbelight]

With @superchromix we noticed that the implementation of COMET in the Colab notebook is different from the one in the main repository, using CLI. It should be possible to import the repository packages inside the Colab notebook using a pip install on the project. Such change would make code maintenance much easier in the long run.

A useful command line for this would be something like that:
!pip install git+https://github.com/gpufit/Comet.git@master#subdirectory=Python_interface

It might require uninstalling a few things from the notebook first to allow for the project to be installed with the specific versions specified in the .toml file in order to work, or allow less restrictive package version list, I am not sure.

[LREIN663]

Wonderful idea!
After doing quite some testing with this, the most straightforward way I found to realize this is to install the package without all the heavy dependencies from the .toml file. With the pre-installed packages on the colab machines it comes down to just installing numba cuda and COMET. I included this with todays merge with the dev version: 4c2d641.

[RomainAbbelight]

Thanks a lot! That sounds great.