fix(heatconduction)!: name the volume target volfrac

[mkeeler43]

Stacked on #280. Merge #280 first, and this PR then retargets to main.

Why

After #280, every volume-constrained problem calls its target volfrac except HeatConduction2D and 3D, which call it volume. This renames it, so one name means the same thing across the benchmark.

What changes

• Code. The condition, the constraint argument, the config keys and the initialize_design keyword change from volume to volfrac in both problems. The constraint has to follow, because it receives its inputs by parameter name.
• Datasets. Both problems point at new dataset repos, heat_conduction_2d_v1 and heat_conduction_3d_v1. They are copies of v0 with one column renamed. Nothing was recomputed. Every other column was checked against the source and is identical, with the same row order.
• Container. The scripts that run inside the container read their inputs by position, not by name, so they need no change.

The problem versions stay at 0, as in #280.

Breaking

Callers pass volfrac instead of volume. That applies to the constructor config, to simulate and optimize configs, and to initialize_design.

An unrecognized config key is ignored, not rejected. So a caller that still passes volume in a simulate config silently gets the default target. Passing it to the constructor fails loudly.

Checks

Pre-commit passes. A design at the default target gives no constraint violation, and one 0.05 above it gives exactly one. The condition order is unchanged, with volfrac first.

🤖 Generated with Claude Code