limited both scheduler to 1 CPU fix

all of

• cpu_count
• cpu_percent
• cpu_shares
• cpu_period
• cpu_quota

seems brittle, heavily dependent on the version and engine and in the end does not limit the resources.

i changed to cpus which seems somewhat stable. I also set the two sheduler to one cpu fix for now. if this become a bottleneck we can loosen the limit.