- Better self-containment: read number of expected samples from output
rather than the environment
- Less reliance on `stdout` timing: Count samples in application and
output it rather then counting flushed dots.
pyterm only echoes after a \n is received, so use socat instead
so '.' are printed as they are generated. This will make the test
timeout for slow rtts like kinetis.
For these slow paltforms also reduce the sample size.
