Clean and isolate model/backend attributes

[irm-codebase]

Fixes #608 and #619

Summary of changes in this pull request

This PR aims to condense the amount of stuff we save throughout the init / build / solve process by avoiding attribute duplication and cleanly defining what is passed to the backend and saved in model files.

• model._timings and model._model_data.attrs["timestamp_xxxx"] are now model.timestamps.
• backend._add_run_mode_math was moved to model._ensure_mode_math.
• model.math / model._model_data.attrs["math"] are now just model.math.
• model.math_documentation was moved to postprocessing.math_documentation, which takes any model as input. This should help users build their own docs without tampering with model.

PR checklist

• rework timestamps
• avoid math additions in the backend
• remove _model_def_dict
• remove unnecessary math documentation attribute
• avoid duplication of math attributes
• ensure backend has limited attribute access

Reviewer checklist

• Test(s) added to cover contribution
• Documentation updated
• Changelog updated
• Coverage maintained or improved

[irm-codebase]

At the moment, use of math.math_documentation is expected to fail. I'll most likely extract this from model, since it is not really needed in a regular case.

[irm-codebase]

model.math_documentation is now its own postprocessing module, and is no longer included in model!
It now receives a model as an input, and build math documentation for it.
Docs are failing because of this, though.

[irm-codebase]

Current implementation is returning a rather funky error when building the documentation for storage_intra_cluster.

For the storage_intra_cluster_max constraint, the method generate_top_level_where_array returns an object with missing dimensions!

To repeat, add a conditional breakpoint in backend_model.py, line 284 with name == "storage_intra_cluster_max".

import calliope
from calliope.postprocess.math_documentation import MathDocumentation


MODEL_PATH = "docs/hooks/dummy_model/model.yaml"
model_config = calliope.AttrDict.from_yaml(MODEL_PATH)

for override in model_config["overrides"].keys():
    custom_model = calliope.Model(model_definition=MODEL_PATH, scenario=override)
    custom_model.build()
    math_documentation = MathDocumentation(custom_model)

Notice for top_level_where.dims

• pyomo backend: ('nodes', 'techs', 'carriers', 'clusters')
• latex backend: ('techs',)

No idea why this happens, yet.

[irm-codebase]

Further debugging:

The error is caused by clusters disappearing from the xarray dataset dimensions????

edit: error was caused by passing model.inputs instead of model._model_data to the latex backend. Attribute filtering also REMOVES dimensions!

[irm-codebase]

Closed in favor of #639