CI: reduce CPU requests in all CI pipelines (#11494)

We are requesting too much CPU per pod running in our CI piplines, which
limits how many CIs can run in parallel due to quota in our namespace

Utilization with old requests is around 5% on average.
Reducing the requests to half should allow running more than 1-2
pipelines at a time

Signed-off-by: NirWolfer <nwolfer@nvidia.com>