fix(ingest): trust CADC over _report.txt when deciding tile status

[Sebokolodi]

Symptom

do_ingest is marking tiles IngestFailed even when CADC contains every expected product. Observed today on tiles 11720 and 11532 (both 943MHz):

check_CADC-0fa  →  CADC contains all products.
prefect.flow_runs →  _report.txt reports that ingestion failed
update_tile_database-3b7  →  Updated tile 11720 status to IngestFailed

A previous run uploaded all 17 expected 3D-pipeline products. A re-run hit a transient minoc HEAD/PUT timeout on one or two files, so the run's _report.txt showed Number of Successes: 23 (or less) / 24. The status decision logic gates Ingested on both _report.txt strict 24/24 and CADC having all products, so the tile got marked IngestFailed even though the data is fully present.

Cause

ingest3Dpipeline.py:319-326 before this PR:

status = "IngestFailed"
if success:                # _report.txt strict 24/24
    if CADCsuccess:
        status = "Ingested"
    else:
        print("_report.txt reports succesful ingest, but CADC ingest failed")
else:
    print("_report.txt reports that ingestion failed")

_report.txt only records the per-file outcomes of the current run. It's not the source of truth for whether the tile is actually fully at CADC — CADC's holdings are.

Fix

Flip the priority: CADC is the source of truth. If all expected products are visible at CADC, the tile is Ingested, regardless of what this particular run's _report.txt says. Only mark IngestFailed when CADC is actually missing products.

status = "IngestFailed"
if CADCsuccess:
    status = "Ingested"
    if not success:
        print("_report.txt reports per-file errors but CADC contains all expected products; treating as Ingested.")
elif success:
    print("_report.txt reports successful ingest but CADC is missing products.")
else:
    print("_report.txt reports that ingestion failed and CADC is missing products.")

What this fixes (and doesn't)

Scenario	Before	After
CADC has all 17 products, _report.txt = 24/24	Ingested	Ingested
CADC has all 17 products, _report.txt < 24/24 (transient timeouts; data was uploaded by an earlier run)	IngestFailed (false negative)	Ingested ✅
CADC missing products, _report.txt < 24/24	IngestFailed	IngestFailed
CADC missing products, _report.txt = 24/24 (very unusual, would indicate something off in caom2pipe)	Ingested (was wrong)	IngestFailed ✅

Tiles affected by the false-negative case so far on v3.0.0: 11720, 11532, 11846, and likely earlier 11293.

Out of scope: tile 11519 has a genuinely missing product (RMSF_real_p3d_v1) from a separate caom2utils blueprint double-configuration bug in the Output3DMapping chain — that's a real data-loss case and will still correctly mark IngestFailed after this change. That bug needs a fix on the possum2caom2 / caom2utils side (raising separately).

Test plan

• Re-run do_ingest on tile 11720 (943MHz). With this fix, it should mark Ingested because CADC already has all 17 products.
• Re-run on tile 11519. Should still mark IngestFailed (correct) — RMSF_real_p3d_v1 is genuinely missing.
• Spot-check the spreadsheet / DB: any tile previously stuck on IngestFailed that is actually fully at CADC will need a manual status correction (this PR only fixes future runs).