Port `dataset dry-run` / `dataset impact` from the Go fork

[beengud]

Port dataset dry-run / dataset impact from the Go fork

Source: beengud/observe PR #5. Files: cmd_dataset.go, ot_dataset.go.
New CLI status: Absent — new CLI dataset only supports list and view.

Old command surface

observe dataset dry-run <file.json>    # dry-run saving a dataset pipeline; nothing created/modified
observe dataset impact <file.json>     # which downstream datasets are affected by the proposed save

Both take the same JSON:

{ "workspaceId": "...", "dataset": { "name": "..." },
  "query": { "stageQueries": [ { "stageID": "main", "pipeline": "filter true" } ] } }

Behavior / API

• dry-run: reports the dataset that would be saved, any datasets that would be dematerialized/rematerialized, and any compile/validation errors. Exit 1 if any error datasets.
• impact: table of affected downstream datasets (name, ID, dependency type); error datasets to stderr.

Porting notes ⚠️

• Same stageID/stageId normalization caveat as board/worksheet applies to stageQueries inputs (old fork null-schema fixes).

Acceptance

• Both subcommands ported with matching output and exit codes; tests ported.