Add per-test max_time override for system tests

[AdityaGupta716]

What this PR does

Adds an optional max_time field to entries in tests.yaml to cap preCICE simulation time per test case.

How it works:

• Reads a max_time value from tests.yaml
• Overwrites the <max-time> (or <max-time-windows>) value in precice-config.xml while preparing the case to run
• Applies consistently to both test runs and reference result generation

Closes #402

Checklist

• I added a summary of any user-facing changes in changelog-entries/402.md
• I will remember to squash-and-merge with a useful summary

[AdityaGupta716]

@MakisH Plz review

[precice-bot]

This pull request has been mentioned on preCICE Forum on Discourse. There might be relevant details there:

https://precice.discourse.group/t/gsoc-2026-aditya-gupta/2773/4

[MakisH]

How does this PR relate to the previously opened #738?

[AdityaGupta716]

@MakisH yes PR tackles the same issue as #738. I opened it separately because I wanted to keep the scope focused just the max_time override, without the iterations-log changes that #738 also includes.
There are a couple of small differences in approach too: this one fails fast on invalid max_time values instead of silently skipping, and adds a README example.

[MakisH]

FYI, this PR (and some other) could benefit from:

• A short explanation of the main idea:
  • Reads a max-time value from tests.yaml
  • Overwrites the value in the precice-config.xml while preparing the case to run.
• Some Markdown formatting, especially where you show code.

Given my previous comment on the relation to another PR (#744 (comment)), note that I still need to review the other PR as well. Maybe don't rush implementing everything.

[MakisH]

I tested this locally for max_time, and it works. Everything else looks good now.

Note that fieldcompare fails due to missing source files. I tried adding --ignore-missing-source-files, but then it still fails in the last time window, since the behavior is a bit different (probably a bug in the OpenFOAM adapter) and gave an error of just above the tolerance. This means that we cannot reuse the same reference results without some handling, but it is still useful for more complex cases.

Thanks again for contributing!