Trying to run static mode without setup (or update) crashes the model

[jacobcook1995]

Describe the bug
If you follow the instructions given in the docs for running a model in static mode and take the approach that is supposed to skip the model setup (or update) process the simulation run crashes during initialisation

To Reproduce
Steps to reproduce the behavior:

1. Set up a virtual_ecosystem example simulation
2. Configure a model (e.g. soil) to be static
3. Provide all variables in that models _vars_populated_by_init as input data
4. Run the simulation
5. See error, i.e. "ValueError: Variable dissolved_nitrate initialised by soil already initialised by ['data']."

Expected behavior
Setup (or update) is skipped

Additional context
The question here is really whether this is a bug in the static model that should be fixed, or just a documentation problem (which was my assumption with #1218). An alternative option to make a model static, is to delete all references from it from the config and instead just provide the outputs that it should generate. This gives us a way to skip the update and setup process, but not the setup process by itself

[davidorme]

This is a bug in the static model. I wondered if I'd broken something in #1296 but I've checked out the last commit to develop preceding merging that PR (9ed4716) and we get the same bug:

(virtual-ecosystem-py3.11) ➜  virtual_ecosystem git:(9ed47167) ✗ cd ~/Downloads/ve_example 
(virtual-ecosystem-py3.11) ➜  ve_example ve_run config --outpath out --logfile out/out.log
Starting Virtual Ecosystem simulation.
Logging to: out/out.log
* Loading configuration
* Saved compiled configuration: out/ve_full_model_configuration.toml
* Built core model components
* Initial data loaded
Traceback (most recent call last):
  File "/Users/dorme/Library/Caches/pypoetry/virtualenvs/virtual-ecosystem-In6MogPy-py3.11/bin/ve_run", line 6, in <module>
    sys.exit(ve_run_cli())
             ^^^^^^^^^^^^
  File "/Users/dorme/Research/Virtual_Rainforest/virtual_ecosystem/virtual_ecosystem/entry_points.py", line 261, in ve_run_cli
    ve_run(
  File "/Users/dorme/Research/Virtual_Rainforest/virtual_ecosystem/virtual_ecosystem/main.py", line 168, in ve_run
    variables.setup_variables(
  File "/Users/dorme/Research/Virtual_Rainforest/virtual_ecosystem/virtual_ecosystem/core/variables.py", line 465, in setup_variables
    _collect_vars_populated_by_init(models)
  File "/Users/dorme/Research/Virtual_Rainforest/virtual_ecosystem/virtual_ecosystem/core/variables.py", line 292, in _collect_vars_populated_by_init
    raise ValueError(
ValueError: Variable dissolved_nitrate initialised by soil already in registry as initialised by ['data'].

So - the "skip both setup and update" option is broken because core.variables._collect_vars_required_for_init does not allow pre-existing variables in data through.

There's a bit of a chicken and egg issue. Ideally, we'd want to tell the variable setup which models are static, so we can handle this explicitly. But because we've model configuration happens in the same step as model setup (__init__ and then _setup) we don't have model configs loaded until after we need the variables resolved. But - we do have the configuration data, so it is probably easiest to simply read out the model static setting for each model and pass this in to variables to try and catch this. [ETA - this is really just moaning about not being able to use model.static and instead having to extract the static settings from the broader config dictionary, which is ugly but not rocket science]

Or we could simply allow __init__ variables to be provided in data, probably with a DEBUG message, and then rely on the model initialisation methods for setting up the static status to handle validation and error reporting.

[jacobcook1995]

As this isn't a MVP feature, we can address this in the short-term by updating the docs to mention that the options don't work (see #1218).

Worth keeping this issue open as this is something that we would want to fix in the long-term