Clean up code & improve test coverage (#249)

* Fix testModelicaSystemLoop

* Use python3 features to clean up ModelicaSystem

- use f-strings to make sendExpression() arguments easier to read,
- use pathlib.Path to convert \ to / in file paths on Windows
- fix a few bugs (mostly missing spaces in error strings)

* Fix warning in test_linearization

* Add more tests for ModelicaSystem methods

* Add test for ModelicaSystem.getLinear*()

* Add test for ModelicaSystem.getContinuous()

It behaves differently before and after simulate()

* Improve test coverage

* Remove unreachable code

* Add test for ModelicaSystem.optimize()

* Use pathlib.Path in OMCSessionHelper

I'm pretty sure this also fixes a bug:
the only Exception I can see os.path.join() raise in _get_omc_path()
is when self.omhome is None, which shouldn't really happen (the
constructor checks it).

I think the intention was to check whether the file actually exists,
so I do that instead.

* Add tests for simulate() edge cases

* Improve test coverage

* Use setCommandLineOptions()

... instead of raw sendExpression.
Seems cleaner and improves test coverage.

* Use f-strings in OMCSessionZMQ

* Simplify omc_process kill logic

... by dropping Python2 support

* Drop Python2 workarounds

* Use f-strings in OMCSessionZMQ

... except for logging calls. It probably doesn't matter, but the
built-in %-formatting does not run when logging is disabled,
improving performance.

* Remove unreachable code

* Fix _get_omc_path fails on Windows

... probably due to missing .exe suffix.

* Fix createCSVData on Windows

The csv module documentation says to open the file with newline='':
https://docs.python.org/3/library/csv.html#id4

Also, I'm pretty sure this isn't how you are supposed to use csv.writer.
delimiter='\n' doesn't seem right.
We should either let the csv writer actually generate the rows, or
remove it altogether and just write the string rows into the file.

* Do not run omc with shell=True

* Remove last remaining str.format() call

Author: ondras12345