Add and end-to-end test runned by a CI to make sure we can always run an experiment

[akouete-kpakpo]

Target

Make sure we can always run an experiment without error

Tasks

• [x] add CI
  • [x] add a .github/workflows/devops-ci.yml to configure the CI job
  • [x] add a Makefile with instructions to
    • [x] build the venv
    • [x] isntall requirements
    • [x] activate venv and run unit tests
• [x] add a unit test for rice irrigation problem

[akouete-kpakpo]

Running python -m nevergrad.benchmark rice_irrigation --num_workers=6 --plot is too long to be integrated in a CI

I have to implement a quicker test but even the simple call to

from nevergrad.functions.irrigation import Irrigation

func = Irrigation(2)

is too long.

I started a profiler and the problem seems to come from the line https://github.com/akouete-kpakpo/nevergrad/blob/e94ba9f0a1ceed9edc2a3d036a3027a933e65cec/nevergrad/functions/irrigation/irrigation.py#L52.

I have to understand why this for loop on 1k samples. Then I can move it to a parameter of the class so that CI would build the function with only 1 sample

[akouete-kpakpo]

make pytest var=nevergrad/functions/irrigation/ mark=ut to launch unit tests on irrigation