Split the slowest smoke test to reduce PR check time

[xiang17]

This smoke test is the only one that takes much longer (twice of most other tasks). When making multiple iterations on many PRs, this task is always the one that prolongs waiting time.

Example of one PR:

job	seconds	minutes
common / CoreAndFilter3xUsingOld3xAgent:CoreAndFilter3xUsingOld3xAgentTest	1057	17.6
common / PreAggMetricsWithRoleNameOverridesAndSampling:PreAggMetricsWithRoleNameOverridesAndSamplingTest	739	12.3
common / RuntimeAttach:RuntimeAttachTest	703	11.7
common / CoreAndFilter3x:CoreAndFilter3xTest	629	10.5
common / SamplingOverridesBackCompat:SamplingOverridesBackCompatTest	521	8.7
common / OpenTelemetryMetric:OpenTelemetryMetricTest	519	8.7