Preserve MediaDive's explicit mixed-sugar solution instead of food-context Sugar

Related: #286, #650, and the already closed MicrobeDecoder-specific #1180.
This is a separate current MediaDive occurrence, not a reopening of the two
MicrobeDecoder record-scoped decisions.

Reproduced current assertion

The accepted isolated MediaDive replay has this full source-bound observation:

• mediadive.solution:4338#recipe/16, media J537.
• Original compound: Sugar, compound_id: 1795.
• Original qualifier: 250 mM each of xylose, maltose and cellobiose.
• Original amount/unit: 0.1 ml; no raw CAS field.
• Current object: NCIT:C71939, with biolink:has_part / BFO:0000051,
  infores:mediadive, observation, manual_agent.

The complete edge's source JSON equals the original recipe object at array
index 15. The source amount and qualifier must not be reassigned to sucrose
or to any single constituent.

The primary MediaDive J537 recipe
independently displays the same mixed-sugar qualifier and quantity. Current
native NCIT projection classifies NCIT:C71939 as Food, and the committed
native-evidence fixture records its sweetener/preservative definition. This
does not establish identity of the named three-sugar solution with that term.
The supported MIM table's generic MIM:Sugar -> NCIT:C71939 assertion remains
unchanged and is evidence to audit, not permission to discard source context.
The original JCM medium 537
also identifies the three-sugar preparation as a solution. Food-context here
describes NCIT's semantics; it is not a regulatory food-grade certification.

Bounded resolution

1. Preserve this explicitly qualified source material on its existing local
   MediaDive identity; do not globally ban Sugar/NCIT:C71939 or rewrite MIM.
2. Ensure the reviewed source qualifier wins over unified, strict/hydrate,
   embedded and direct-call fallback routes. Never borrow a qualifier from a
   different occurrence with the same compound ID.
3. Retain complete original candidate claims and source JSON in the existing
   material-scope audit, with native input bindings and physical multiplicity.
4. Add hermetic positive/negative scope and fallback tests; keep raw quantity,
   assertion ordinal, provenance and unrelated food-context Sugar uses intact.
5. Verify actual source replay and final source-to-merged projection. A label
   join, passing unit tests or a successful merge is not release acceptance.

Do not infer independent utilization of the three constituents or generate
new component assertions as part of this identity repair.

Reproducibility

• Saved raw solutions SHA-256:
  9ab0fbf72c2e5ac25ecab3ada268dfe524db971daa86d67532dea6a5412a5c44.
• Accepted W4 source edges SHA-256:
  05ecb2df49afa66607393fbc0d6f9e15980315ba7bd088d38793d5c329fbf787,
  physical line 95,389.
• Source finalization SHA-256:
  bb505dd2eb022b1a1d7036e5f22b2c62c78acc43627640f71ff30af8f9329f1f.
• Read-only display-join evidence under
  data/issue1224-quarantine-20260929.xjuS9H/issue286-display-join-mo25m5zm/;
  groups.json SHA-256
  56d0c510b84106de7a124f927f4428bab95a3879593bdca970424093e7092ec4.
  The join retained all 5,473 selected raw occurrence candidates and their full
  source rows. It does not validate historical CAS assignments or chemical
  equivalence. This specific defect is established from the actual raw/edge
  witness, not from display spelling alone.

PR #1239 and the separate #1241/#1242 candidate remain frozen for their own
review; this finding has not been silently folded into those accepted deltas.