One client per actor

[anschweitzer]

The key changes are in actor_base.py, scada_base.py and cloud_base.py. The tests pass and a quick manual tests appeared to show that comm was restored after restarting mostquitto broker.

Some defensive coding was added to tests:

1. Most have try/finally blocks so that the actors can be stopped even when an assertion fails the test. We (I) should make this more automatic in the future, but this is enough for the moment.
2. I replaced some (most?) sleeps with calls to wait_for().
3. I commented out test_atn_cli() because it was failing in a flaky way in CI and sometimes locally.
4. Possibly related to 3, I wonder if the BooleanActuator sensing loop could use similar treatment to what we just did to SimpleSensor's loop? I'm not sure I'm following the code, if it looks fine to you I probably just am not absorbing it.

[codecov[bot]]

Codecov Report

Merging #75 (531e390) into main (f8f16f6) will increase coverage by 4.61%. The diff coverage is 91.81%.

@@            Coverage Diff             @@
##             main      #75      +/-   ##
==========================================
+ Coverage   71.43%   76.05%   +4.61%     
==========================================
  Files         134      135       +1     
  Lines        4481     4522      +41     
==========================================
+ Hits         3201     3439     +238     
+ Misses       1280     1083     -197     

Impacted Files	Coverage Δ
test/utils.py	72.22% <72.22%> (ø)
test/test_examples.py	95.25% <92.08%> (-0.42%)	:arrow_down:
gw_spaceheat/actors/cloud_base.py	89.36% <94.44%> (+10.31%)	:arrow_up:
gw_spaceheat/actors/actor_base.py	91.91% <95.83%> (+10.72%)	:arrow_up:
gw_spaceheat/actors/scada_base.py	89.70% <100.00%> (+12.32%)	:arrow_up:
gw_spaceheat/actors/simple_sensor.py	95.91% <0.00%> (-2.05%)	:arrow_down:
gw_spaceheat/schema/property_format.py	54.79% <0.00%> (+8.21%)	:arrow_up:
... and 17 more

:mega: Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

[anschweitzer]

test_async_power_metering_dag() remains flaky. It's unclear to me if is worse on in the PR.

[jessicamillar]

The BooleanActuator reporting is a little different than the SimpleSensor, because it reports on change as well as synchronously. Also, when the reporting.SamplePeriodS is equal to the reporting.ReportingPeriodS I want to offset the timing so that the scada isn't flushing its latest readings exactly at the same time it is getting a synchronous report. This issue will go away if we have a single actor. But I'll clean up the BooleanActuator and the pending PowerMeter so that what they are doing is easier to test and also more clear. In a different PR. (And also dig into the test_async_power_metering_dag() flaikiness)