fix(data): merge 651 confirmed monitor near-duplicate records

[Seungpyo1007]

Continuing the "near-duplicates in other categories not yet scanned" item after the tablet/watch (#324), smartphone (#329) and laptop (#332) passes, I scanned the monitor category. Scope: 882 data/monitor/<brand>/<year>/<base_model_slug>/<slug>.json records, all sourced from one Kaggle Amazon monitor dataset (amazon-monitor-durjoychandrapaul).

Result: 651 confirmed near-duplicate records merged. 882 → 231.

Method

Same approach as the prior passes: normalize slugs/names (punctuation- and brand-prefix-agnostic) and cross-check the specs that must be identical for the same physical panel — release_date, size_inch, resolution, aspect_ratio, refresh_hz, panel_type, ppi, plus the model number embedded in name. The deterministic merge rule: group by base_model_slug + a full spec fingerprint (every field except variant, slug, source_urls); for each cluster keep the lowest source_row as canonical, union source_urls, and delete the rest.

Per-pattern results

• Pattern 1 (same scrape imported twice under different row-id slugs): 650 files merged across 14 base models. Each cluster is byte-identical except slug and variant.source_row. Kept the lowest source_row, deleted the redundant files; source_urls were identical so the union was a no-op and no values changed. Breakdown:
  • asus-va329he 84→1 (−83)
  • aopen-aopen-by-acer (32SA2QK 4K) 69→1 (−68)
  • acer-cb272 63→1 (−62)
  • acer-sh242y 60→1 (−59)
  • acer-xb273k 60→1 (−59)
  • acer-xz342cu 47→1 (−46)
  • acer-kb252q 45→1 (−44)
  • acer-sb240y (G0bi) 44→1 (−43)
  • acer-sb272 (EBI) 39→1 (−38)
  • acer-xz320qu 31→1 (−30)
  • acer-hdr10 (31.5" Nitro 240Hz, two byte-identical price clusters) 74→2 (−72)
  • acer-sb270 27→1 (−26)
  • acer-kc242y 14→1 (−13)
  • acer-xb273u 7→1 (−6)
• Pattern 2 (brand-prefix path variance, e.g. dell-u2723qe vs u2723qe): none. Every record already lives under a consistent brand-prefixed base dir.
• Pattern 3 (cross-source dup under a different title convention): 1 merged. asus-va24ehe rows 60 and 404 are the same ASUS VA24EHE (23.8", 1920×1080, 75 Hz, IPS, $99) imported with two different Amazon title strings. Kept row 60 (cleaner, untruncated title) and adopted the more complete features.adaptive_sync = "FreeSync" from row 404 — a real, documented VA24EHE spec (ASUS product page). No other field differed.

Left unmerged (genuine distinct SKUs / ambiguous — documented, not guessed)

These share a coarse base_model_slug but are different real products, so they were intentionally kept:

• acer-eda343cur — 100 Hz/$200 vs 180 Hz/$280 (different refresh SKUs)
• acer-sb240y — G0bi (black, $95) vs G0wi (white, $110)
• acer-sb242y — EBI (IPS) vs Hbi ($128)
• acer-sb272 — EBI (new) vs "E … (Renewed)" listing
• aopen-aopen-by-acer — 32SA2QK (31.5" 4K) vs 27E1 (27" FHD)
• gigabyte-m27q — M27Q vs M27Q PRO
• hp-hp-series-monitor — Series 5 24" vs 524sa bundle
• lenovo-lenovo-thinkvision-full — ThinkVision M15 (15.6") vs T22i-30 (21.5")
• lg-lg-digital-signage — 49UH5F-H (49") vs 86UH5E-B (86")
• lg-lg-ultragear-gaming — 27" QHD 240 Hz vs 27" 5120×1440 165 Hz
• lilliput-fa1014-np — 1280×800 non-touch vs 1920×1080 touch
• samsung-hdr2000 — Odyssey Neo G7 vs Neo G8
• acer-hdr10 residual — the hdr10 slug is an importer fallback shared by a few genuinely different Nitros (27" WQHD 170 Hz, 27" FHD 240 Hz, and the 31.5" 240 Hz at two prices); left as-is rather than guess.

Placeholder / fabricated-defaults check: clean (no bug)

No equivalent of the phone-batch battery_mah/weight_g stamping. Spec value distributions across distinct base models are organic (27" popular; 1920×1080 common; 100/165/75 Hz common), each spread over many unrelated models — not one value stamped onto unrelated devices. The one systematic artifact is release_date = "2023-01-01" on all 882 records, but that is a uniform dataset-wide import placeholder that affects every record equally, so it is not a dedup signal and is out of scope for this merge. Noting it here for the "placeholder dates" tracker item.

Verification

• python -m app.validate → Data validation passed (monitor required-field / slug-convention / filename-match / base-dir checks included).
• python integrity_check.py <data> --strict → no hard anomalies (structural section clean: no duplicate monitor slugs, no slug/filename mismatches, no verified-without-source); only the pre-existing advisory CPU/GPU ratio outliers print, which the gate treats as review-only.
• Dump refreshed as the PR's last commit (python -m app.dump), scoped to the monitor collection: the 651 orphaned detail pages were pruned, v1/monitors/index.json rebuilt (231), the asus-va24ehe detail page picked up the merged adaptive_sync, and the v1 manifest monitor count updated 882 → 231. Unrelated pre-existing serialization drift in other collections was deliberately kept out to keep this diff monitor-scoped and reviewable.

Closes #296