With the recent merge of a plugin that was previously in a external repository and overall the increase of the size of the codebase running the unit tests with the coverage enabled in the test pipeline requires a not negligible amount of memory. With the current size of the machines used as agents to run the pipeline we sometimes ending up OOM which result in a build failure :/ .

Since:
* there is no required code coverage enforcement at this time
* probably nobody looks at the code coverage of each builds
* we would need bigger build agents to "solve" the issue
it's probably fine to only collect the coverage in the nightly pipeline. This would also make the overall pipeline a bit faster.