add cell_division_direction_function for divide()

[rheiland]

• add a cell custom function as an alternative to randomly positioning the daughter cell in divide()
• if this function is defined (non-NULL), it will be called and return a 3-vector which is the normalized direction from the parent cell to the daughter cell
• this PR does not define another sample project to test/demo; rather, I'd like to suggest that we consider adopting a .zip approach that bundles all necessary files for compiling and testing. For now, we could simply adopt the user_projects workflow.

To test:

• clone this branch, then:

make
make reset
# download https://github.com/rheiland/sample_models/blob/main/bacterial_growth.zip
# and put into `/user_projects` and  unzip it there, then:
make load PROJ=bacterial_growth
make
project    # or use the Studio to run & plot results

Sample output:

[drbergman]

This is actually quite timely and really useful! I think a couple minor changes will improve on this in ways worth considering though. Looking forward to discussing further!

[drbergman]

any reason not to initialize cell_division_direction_function as cell_division_orientation? simplifies the logic here? I know we follow this structure you have here elsewhere in the codebase, but I think setting the default and overriding is simpler and yields simpler logic here.

[rheiland]

Thanks for the feedback! Offhand, I'm OK with both of your suggestions: init cell_division_direction_function ascell_division_orientation, and normalizing the vector returned by the custom function (even if they already did) - a safer approach. I'll sit on it awhile in case others chime in. I'm also curious to hear if I've missed something more fundamental, as this topic came up in a Macklin lab mtg recently.

[drbergman]

I'm curious to hear as well...the only other thing that comes to mind is managing the final rand_vec orientation in a 2D sim to be in the xy-plane.

[rheiland]

Actually, that might be one argument for keeping the if-else logic on the cell_division_direction_function in divide(): we could explicitly check the user-supplied vector for being in the x-y plane for a 2D sim (which implies additional checks for 2D/3D sim and z ~= 0.0). Ugh.

[drbergman]

we could write a setter function that sends a default cell through the function assert these condition(s) hold. Normalization regardless. in xy-plane when in 2D.