CI development

This MR will include many useful unit tests in the continuous integration. This long overdue upgrade will save a lot of manual effort in making plots for reviews, reduce the occurrence of bugs in the code, and make bugs easier to catch right away. Here are the tests that are now included:

• Run 3 versions of the calibration pipeline, and check that the right number of calibrated frames are produced
• Check the ASDs of GDS-CALIB_STRAIN and GDS-CALIB_STRAIN_CLEAN, and compare them to a "standard", nominally correct ASD. Plots are included in the CI job artifacts.
• Check the TDCFs, and compare them to nominally correct TDCFs. Plots included.
• Check the calibration state vector, and compare each bit to a nominally correct state vector. Plots included.
• Measure ratios of h(t) / Pcal and h(t) / Actuation injections, and compare to a nominally correct version. Time-series plots included.
• Use FrDiff to compare GDS-CALIB_STRAIN_CLEAN in frames produced by two otherwise identical pipelines with different start times. They are required to be identical.
• Measure the latency in a simulated low-latency-like configuration. Check that the mean latency is not too large, that there are not frequent or large excursions, and that the latency does not increase steadily with time.

Job artifacts from the most recent commit (at the time of writing) can be seen here, including plots and txt files with the data being plotted.

Getting all stages of the CI to work also required me to address some packaging issues. If I did things correctly, this MR Closes #7 (closed), #12 (closed), #14 (closed), #15 (closed), #26 (closed). I was not able to add @duncanmmacleod as a reviewer, but he created several of these issues.

tests/tests_pytest/error.py

@@ -143,5 +143,5 @@ def rms(d_type, typ='A'):
         e_file.close()
         print(s2, file = sys.stderr)
 
-        assert rms < 0.001, "%s: error too large!" % name[i]
+        assert rms < 0.002, "%s: error too large!" % name[i]