Fix bug defining subsector commodities when zero existing capacity

[tsmbland]

Description

MUSE uses subsectors which are in charge of one or more commodities. Users can explicitly pass a list of commodities in the settings file, or MUSE can infer the commodities based on the ExistingCapacity file.

However, there's a problem with how this is done. MUSE first creates assets from the ExistingCapacity, then aggregates the end-use commodities from these assets. However, if there are technologies in ExistingCapacity with zero initial capacity, no assets will be created for these, and these won't be included in the end-use aggregation process.

E.g. a refineries sector may be in charge of oil and hydrogen. If there's no initial demand for hydrogen, the user will set initial capacity for hydrogen-producing technologies to zero. However, demand for hydrogen may increase in the future, so we still want to include hydrogen in the subsector's self.commodities so that hydrogen demand get's passed to the subsector.

I've fixed this by aggregating the enduse commodities from ExistingCapacity directly.

I've added another step during investment to remove commodities with zero demand. E.g in the above example we wouldn't want it to perform investments to meet hydrogen demand when demand is zero (there will be no investments, but it's just a waste of computation). This is probably a good thing to do anyway, but seemed especially appropriate now.

Fixes # (issue)

Type of change

• New feature (non-breaking change which adds functionality)
• Optimization (non-breaking, back-end change that speeds up the code)
• Bug fix (non-breaking change which fixes an issue)
• Breaking change (whatever its nature)

Key checklist

• All tests pass: $ python -m pytest
• The documentation builds and looks OK: $ python -m sphinx -b html docs docs/build

Further checks

• Code is commented, particularly in hard-to-understand areas
• Tests added that prove fix is effective or that feature works

[Copilot]

Pull Request Overview

This PR fixes a bug where subsector commodities defined from the ExistingCapacity file were missing when their initial capacity was zero. The changes update the aggregation logic in the subsector and related tests, and adjust agent factory inputs for consistency.

• Removed the use of agent.assets in aggregate_enduses and now infer commodities directly from technologies.
• Updated the agent factory to require parsed capacity data.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
tests/test_subsector.py	Updated tests to call aggregate_enduses with technologies only.
src/muse/sectors/subsector.py	Revised commodity aggregation and demand filtering; now aggregates directly from technologies.
src/muse/agents/factories.py	Refined the capacity parameter type to be a DataArray (removing support for file paths).

[Copilot]

Since capacity is now strictly expected to be a DataArray, consider updating related documentation and usages to make it clear that file paths are no longer supported.

[dalonsoa]

LGTM! What an impact for such small change. I guess that none of the examples and tutorials were affected by this, right?

[tsmbland]

LGTM! What an impact for such small change. I guess that none of the examples and tutorials were affected by this, right?

Doesn't affect any of the examples, but was picked up by a researcher using a more complicated model. (In fact this is a bug that I accidentally introduced in v1.4, which is why it's only being picked up now 🫢)