These tests are expensive to run and do not give meaningful coverageinformation for the flutter tool (since they are black-box tests thatrun the tool as a subprocess, rather than being unit tests). For thisreason, they are in a separate shard when running on continuousintegration and are not run when calculating coverage.