Revisit the CI triggers for cpp jobs

[rdmark]

As has been discussed back and forth in the past, we currently have CI job triggers for both pushes and pull requests. This leads to duplicate jobs every time a new commit is pushed to an active pull request.

We're on the free tier of GitHub for Teams, so computing resources are theoretically not unlimited (2000 actions minutes monthly) although I've never seen us exceed the limits yet. Our cpp jobs are quite heavy, with compilation, unit tests, and Sonar static analysis of the entire codebase (including python). A recent PR used up ~70min of computing time for each push, and with three pushes I theoretically used up 210 min of computing time, over 1/10 of the monthly allowance.

Off the top of my head, if we are already running jobs for each push to remote, do we really need the PR triggers at all?

[rdmark]

Addressed as part of #1417