chore(examples): dump db_id pyramid in simple_nft_dual to aid #160 triage

[kalwalt]

Refs #160. Small follow-up to #159 (merged).

Summary

simple_nft_dual.rs now prints all reference-image pyramid levels read from the .fset3 right after feed_ref_data succeeds:

Reference pyramid (9 levels):
  db_id=0 -> 893x1117 px
  db_id=1 -> 750x938 px
  db_id=2 -> 595x745 px
  db_id=3 -> 473x591 px
  db_id=4 -> 375x469 px
  db_id=5 -> 298x372 px
  db_id=6 -> 236x296 px
  db_id=7 -> 188x235 px
  db_id=8 -> 149x186 px

Pure diagnostic. No behavioural change.

Why

When the dual matcher reports matched_id = 2 and a tier-2 divergence at "ref image 595x745", anyone reproducing #160 can now immediately see which db_id got matched, the full set of pyramid levels available, and confirm that 595x745 is a legitimate ~2/3-scale level of the 893x1117 master (and not a stale-cache or axis-swap pathology). Saves an ad-hoc print step during triage.

The pyramid ratios on pinball are a clean 2^(-1/3) geometric series — i.e. a 1/3-octave pyramid with 3 levels per octave. Useful piece of ground truth for #160's "is this expected BHC-variance vs. a per-scale parity bug?" decision.

Test plan

• cargo fmt --all -- --check clean
• cargo build --features "dual-mode log-helpers" --example simple_nft_dual succeeds
• cargo run --features "dual-mode log-helpers" --example simple_nft_dual produces the dump in the expected position (Phase A, immediately after "Reference data loaded into both backends")

Refs

• Refs investigate: DualFreakMatcher tier-2 divergence on pinball-demo at db_id=2 (~14 px > 2.0 px tolerance) #160 (the issue this aids)
• Related to feat(examples): add simple_nft_dual.rs with DualFreakMatcher and per-frame divergence reporting (#157) #159 (merged — landed the example this enhances)

🤖 Generated with Claude Code