Fix time accumulation and max time

[fsimonis]

Main changes of this PR

This PR

• replaces the time window start with a Kahan aggregator
• constrains the max dt by max-time

Motivation and additional information

time-window-size = 0.01 and max-time = 5 leads to 501 time windows due to rounding errors.
This should solve the problem.

Author's checklist

• I used the pre-commit hook to prevent dirty commits and used pre-commit run --all to format old commits.
• I added a changelog file with make changelog if there are user-observable changes since the last release.
• I added a test to cover the proposed changes in our test suite.
• For breaking changes: I documented the changes in the appropriate porting guide.
• I sticked to C++17 features.
• I sticked to CMake version 3.16.3.
• I squashed / am about to squash all commits that should be seen as one.

[davidscn]

Do we really want to forward here an arbitrary small dt? I would say that the current behavior is numerically still more robust. Not sure that this PR is the right place to add it into the library.

[fsimonis]

The previous state was flatout wrong.
time-window-size = 0.3 and max-time = 1 resulted in a timestep from 0.9 to 1.2.

Now the final time-window-size is correctly synchronized.

[davidscn]

With participant-first methods and adaptive time-stepping (e.g. in OpenFOAM) this can still lead to issues. In the adapter, we set the end time to infinity such that preCICE steers it. Hence, the solver is not aware of the maxTime limit and this return value can lead o arbitrary small dt causing numerical issues then.

[davidscn]

I just realized that this only applies to hasTimeWindowSize(). What happens to the second participant in a participant-first time-stepping. Do we then perform different time-step length for different participants?

[fsimonis]

For participant-first there are two cases:

• first doesn't have a time window size as it is dictating it. The getNextTimeStepMaxSize() is maxTime if provided and inf otherwise.
  This was correctly implemented.
• second has a dynamic time window size which it received from first, which is correctly limited by max-time. So in theory this cannot surpass max-time.