MBDef 1.0.0: provenance without internal ids or machine details

Rebuilt with `taxonomy_pipeline dataset build`: the same tables and Explorer data; provenance.json names the spelling rule in words and records no machine details.

provenance.json

@@ -9,25 +9,21 @@
   "ground_truth_repo_id": "mediabiasgroup/MBDef-Gold"
  },
  "code": {
-  "commit": "da35312d8cca0a08ad9b19089a3e10a5f08bc127",
+  "commit": "d8c069c97a9c1c14a1b839f40e6f9226cc587928",
   "dirty": false,
   "python": "3.14.8",
   "polars": "2.0.0",
-  "pyarrow": "26.0.0",
-  "platform": "Linux-7.3.0-4-cachyos-x86_64-with-glibc2.44"
+  "pyarrow": "26.0.0"
  },
  "decisions": {
   "text": "every definition text word for word",
   "context": "no paragraph context or other text of the papers",
   "rows": "only the kept term-in-paper rows",
-  "concepts": "spellings of a term merged into one concept (label groups), over all corpus terms, kept rows only",
+  "concepts": "spellings of a term merged into one term, over all corpus terms, kept rows only",
   "ground_truth": "a separate dataset, MBDef-Gold",
   "licence": "cc-by-4.0"
  },
- "rule": {
-  "name": "Label Match",
-  "version": "C-20261021"
- },
+ "rule": "two spellings are one term when they are the same after case, hyphens, spacing, quotation marks, a leading article, stop words between content words, a bracketed acronym, word order, British spelling and plurals are set aside (the card's \"Term merging\")",
  "inputs": [
   {
    "path": "$DATA_ROOT/frozen/step5-post-processing/20261021/decisions/ensemble.csv",