Dakota issue 12: support finite-data base surrogate workflows

[JavierGOrdonnez]

Parent epic: #12
Source: ITISFoundation/mmux_python#3

Current status: Partial. Public workflows fit surrogates from finite tabular data, but there is no callable-or-CSV factory/base-surrogate workflow matching the source user story.

Acceptance criteria:

• Define a stable factory or fit entrypoint for callable and finite-data sources.
• Keep CSV/file conversion outside the core API boundary or document the adapter.
• Add integration coverage for a finite dataset with repeated queries.

[JavierGOrdonnez]

Assessment 2026-08-19 — partially addressed (status:in-progress)

The source issue asks for both base options: use a full function/study as the base, or build a base surrogate from user-supplied finite data.

Done. The finite-data path is the only thing itis-sumo does, and it does it well — every study is a pure data-fit surrogate over a caller-supplied table via import_build_points_file, with no interface block at all. A user with nothing but a CSV is fully served today.

Missing. The callable/full-study path does not exist. There is no way to hand itis-sumo a simulator to evaluate, so the "use a full function as base" half of the requirement is unimplemented. This is the same gap that makes #27 (batch vs non-batch callbacks) currently moot.

Next step. Decide whether a callable-driven study belongs in itis-sumo at all, or stays on the consumer side. If it belongs here, it should be specced before implementation since it pulls in the callback-mode question with it.