Add configurable trajectory persistence and decouple thermo logging in melt-quench pipeline

[Gitdowski]

This PR (and its major reviews) improves melt-quench persistence and observability. It also seperates the NVT simulation required for structure analysis averaging if n_averaging_frames > 1 into the structure_analysis workflow.

• Fixed n_averaging_frames plumbing so user-provided values are respected end-to-end.
• Split n_print into n_dump and n_print_thermo to decouple thermo log verbosity from sparse dump cadence.
• Updated also in docs and notebooks
• Kept final_structure behavior unchanged for downstream analyses.
• Improved database docstrings to clearly describe the trajectory dataflow and persistence policy.
• Adding a trajectory_storing_mode independently for for melt-quench and structure analysis workflow to control what part of the trajectory/dump data enters the database:
  • all_frames_all_data keeps all data as recorded
  • all_frames_drop_velocities_and_forces keeps all dumped frames but removes potentially data-heavy forces and velocities
  • last_frame_all_data reduces dump-related data (key-names in parsed_output: natoms, cells, indices, forces, velocities, unwrapped_positions, positions) to only the last recorded frame
  • last_frame_drop_velocities_and_forces
• Note that the trajectory_storing_mode is currently applied individually on all stages of the melt-quenching procedure. Therefore, if the last_frame option is selected, the simulation history will contain the last structure of every stage.
• pixi: set lammpsparser version to >=0.0.7 to have access to new dump_final_structure parameter
• unification of _run_lammps_md function to avoid requiring multiple definitions of it
• Moving the additional NVT run of the melt quenching simulation (required for structure analysis) to the actual structure analysis workflow

[codecov]

Codecov Report

❌ Patch coverage is 82.40741% with 38 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
amorphouspy_api/src/amorphouspy_api/database.py	66.66%	15 Missing ⚠️
...ouspy/src/amorphouspy/properties/structural/all.py	80.00%	10 Missing ⚠️
...phouspy/src/amorphouspy/properties/cte_analysis.py	65.21%	8 Missing ⚠️
amorphouspy/src/amorphouspy/lammps/runner.py	86.95%	3 Missing ⚠️
...morphouspy/src/amorphouspy/properties/viscosity.py	92.85%	1 Missing ⚠️
...py_api/src/amorphouspy_api/routers/jobs_helpers.py	85.71%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[ltalirz]

Thanks for the PR! Happy to merge after the minor changes discussed

[Gitdowski]

Working on a more unified version to avoid seperate definitions of the _run_lammps_md. Expect a major update of this PR today, also taking into account new option in lammpsparser that makes sure the last frame is dumped.

[Gitdowski]

Hey @Atilaac and @ltalirz,
unifying the _run_lammps_md function (now with individual n_dump and n_print_thermo, also now accepting pressure, and making sure the last run of a simulation is always dumped via dump_final_structure = True) had quite some effect on various parts of the package. I tried to consistently use the new parameters everywhere, also on the api side, to avoid confusion.

I also updated the notebooks and copied the changed notebooks to docs/notebooks (@Atilaac: I guess this now also requires the builddocs label ? also any other?)

Please have again a look at the changes.

PS: I am out of office tomorrow. We can discuss today or next week.

[ltalirz]

Let's make this more explicit

persist_trajectory:

• "all":
• "last_frame"

and be explicit that we just store positions and cells (note: the prepare_for_history function needs to be adapted to also drop forces/velocities in case all)

[ltalirz]

@Gitdowski You may also be able to close #340 once this is merged