Planner: keep-in-stock goals & capital (building) material requirements

[ApocDev]

Two linked capabilities that together let a plan account for things it needs but that no recipe in the chain directly consumes — the classic case being "Py Science 1 needs steel" only because the buildings that run the chain are made of steel.

1. "Keep N in stock" goal

A goal type that means "make sure this good exists / keep a buffer of it" rather than "produce it at X/s forever". Today every goal is a continuous throughput target (BlockData.goals: { name, rate }[]); a one-off or buffer item has to be faked as a tiny rate. A stock goal should:

• declare a target quantity (e.g. 100 steel on hand), not a per-second rate;
• expand the recipe chain needed to make it (so its intermediates are covered) without implying a meaningful continuous production rate;
• be visually distinct from rate goals in the block editor and summaries.

2. Capital / build-cost requirements

The machines a plan requires are themselves built from materials, and those materials must be produced even when nothing in the production recipes consumes them. Example: a Py Science 1 block uses assemblers/furnaces/etc. that are made of steel, so steel is genuinely required to stand the factory up — but it never appears as an ingredient of the science chain, so the planner currently shows zero steel demand.

Desired: roll the build cost of the buildings a plan needs into its demands as a one-time / coverage requirement. The solver already computes per-recipe building counts (machineReqs in app/src/server/factorio.ts); this feature would take those counts, look up each building's own item recipe, and surface the materials (and their sub-chain) as required — so steel, circuits, etc. show up as things to build.

Open questions

• Stock and capital requirements are "coverage" demands (expand the chain, no continuous rate) vs the existing throughput goals. Do they share a goal mechanism or stay separate? See the coverage angle raised in Planner: support rate windows and low-throughput stock goals #10 (which otherwise covers rate windows/units — this issue takes over the stock/coverage half of it).
• How deep does capital expansion go — just the building's direct materials, or the full sub-chain (steel → iron → ore)? Probably the full chain, surfaced as demands the factory view can satisfy with other blocks.
• Buildings get replaced/upgraded over time; is capital a one-time cost, or amortized? Start with one-time "to stand this up you need X".
• How do coverage demands interact with the factory-level coherence/seam model and "build the supplier block" flows?

Part of #31.