Implement Access and Stream Trajectories with MDAnalysis inside MolecularNode

Issue Description:

We are planning to enhance the capabilities of MolecularNodes by integrating the ability to access and stream trajectories using MDAnalysis. This feature would enable users to

efficiently load molecular dynamics trajectories without dumping all the frames into memory.
leverage the power of Blender as a live viewer for trajectories e.g. inside a jupyter notebook with BNotebooks Blender kernel.
pave the way for scripting rendering pipelines.

Proposed Design:

To implement this feature, we propose the following design in scripting:

start a session

mda_session = mn.mda.MDAnalysisSession(world_scale=0.01)

show atoms

universe = mda.Universe(top, traj)

custom_selections = {'name_ca': 'name CA', 'ion': 'resname SOD CLA'}

# This will create three representations inside Blender
mda_session.show(universe, name='protein', selection='protein', custom_selections=custom_selections)

# Add new representations for new selected atoms
ligand = universe.select_atoms('resname LIG')
mda_session.show(ligand, name='ligand', representation='ball+stick')

# Add representations for dynamically selected selected atoms
around_ligand = universe.select_atoms('protein and byres (around 4 resname LIG)', updating=True)
mda_session.show(around_ligand, name='around_lig', representation='vdw')

# add new trajectory and an offset of the frame can be set
universe2 = mda.Universe(top2, traj2)
mda_session.show(universe2, name='protein', selection='protein', frame_offset=100)

Persistence

# saving
file_path = "blender_and_mda.blend"
bpy.ops.wm.save_as_mainfile(filepath=file_path)

# loading
bpy.ops.wm.open_mainfile(filepath=file_path)
mda_session = bpy.context.scene.mda_session
universe = mda_session.universe

# persistent even when trajectory files are deleted
mda_session.transfer_to_memory(start, step, end)

# or use legacy code with Blender UI

Example:

showcase: https://drive.google.com/file/d/1A_WlN8oUaMqjV6MsKP4F2lSvtRKa6YVt/view?usp=sharing
notebook: https://gist.github.com/yuxuanzhuang/25aa5a1588c7e19efbfc1b7275718b58
Expected Benefits:
a consistent python API to interact between Blender and MDAnalysis
better organization of atoms representations
Memory efficiency
On-the-fly transformations e.g. fixing PBC, designed rotation, smoothing (position averaging).

Expected Caveats:

the current implementation of blender kernel will open blender automatically in the initialization stage. Ideally, it should be only after a new MDAnalysisSession is started.
Need to have a consistent connection to the trajectory in the session.
During the animation, the state of the Universe will be altered. A racing condition is likely to get induced when the trajectory frames are iterated in the notebook simultaneously.
MDAnalysisSession (together with Universe) will be dumped into a tmp location during saving .blend file i.e. ~/.blender_mda_session/ by default.
a relative path of trajectory is saved for Universe -> where does Blender start matters!
~~potentially create conflicts with the current animation engine and the selection process within the geometric node.~~

Tasks (task will not be implemented in this PR):

[x] Design and implement MDAnalysisSession and AtomGroupInBlender.
[x] Loading and Saving (Persistence)
[x] Deletion
[x] Start with different representations
[ ] ~~Function to switch representations~~
[x] Atomgroup with variable n_atoms i.e. a dynamic selection of atoms
[ ] ~~Update attributes if the topology is modified~~
[x] Multiple universes
[ ] ~~ways to add new time-evolved attributes and display in blender (e.g. occupancy)~~
[ ] ~~Benchmark on streaming~~
[x] Write unit tests to ensure reliability and stability.
[x] Update documentation/tutorial to reflect the new feature and its usage.

Future direction

visualize analysis results e.g. simple distances, angles, etc.

BradyAJohnston / MolecularNodes

Implement Access and Stream Trajectories with MDAnalysis inside MolecularNode #292