refactor handling of orphaned resources in the coordinator

[jluebbe]

Labgrid currently uses bidirectional streams between coordinator and
client/exporter. For the client side this is a good fit, since the
client sends requests and the coordinator can answer directly. However
for the exporter we have a case where nested calls were used in the old
crossbar infrastructure, namely when re-acquiring a resource after the
exporter was offline but the place was kept acquired. We call these
orphaned resources. They replace the real resource on the coordinator
side until the resource can be reacquired on the respective exporter
after it has restarted.

With crossbar, when seeing the resource update, the coordinator could
directly call the exporter to acquire the resource for the specific
place. This was possible since crossbar did the RPC route handling and
arbitrary services connected to the crossbar could provide RPC calls to
the service. With GRPC, we are more constrained. Since we only have a
single Input/Output stream which needs to multiplex different objects,
nested calls are not directly supported, since the exporter side would
still wait for the coordinator to answer its own request.

A different approach to orphaned resource handling is required. The
coordinator now uses a loop where it checks the orphaned resources and
tries to reacquire them if the exporter reappears. This however
introduces another problem, the exporter can be under high load and thus
the acquire request from the coordinator can time out. In this case, we
need to abort the acquisition during a regular lock and in case of an
orphaned resource need to replace the orphaned resource with the
eventually acquired resource from the exporter.

We also need to handle the case where the exporter has an acquired
resource, but the place has been released in the meantime (perhaps due
to a timeout on a normal place acquire), the same poll loop handles this
in the coordinator as well.

All in all this means that the resource acquired state for each place is
not necessarily consistent on the coordinator, but will reach an
eventual consistent state. This should be sufficient, since exporter
restarts with orphaned resources should be relatively rare.

[jluebbe]

I've fixed the error detected by pylint.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 90 lines in your changes missing coverage. Please review.

Project coverage is 55.8%. Comparing base (c609d2d) to head (109ef0d).
Report is 5 commits behind head on master.

✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
labgrid/remote/coordinator.py	0.0%	90 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##           master   #1580     +/-   ##
========================================
- Coverage    56.0%   55.8%   -0.3%     
========================================
  Files         170     170             
  Lines       13307   13365     +58     
========================================
  Hits         7458    7458             
- Misses       5849    5907     +58     

Flag	Coverage Δ
3.10	55.7% <0.0%> (-0.3%)	⬇️
3.11	55.7% <0.0%> (-0.3%)	⬇️
3.12	55.7% <0.0%> (-0.3%)	⬇️
3.13	55.7% <0.0%> (-0.3%)	⬇️
3.9	55.8% <0.0%> (-0.3%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.