…t IFD 1

Two Aperio-ecosystem conformance bugs in v0.24.0 SVS/TIFF output that
OpenSlide/ImageScope reject but opentile (lenient) masked.

1. Edge-tile padding. The retile engine encoded partial right/bottom edge
   tiles — and whole levels smaller than one tile — at truncated content size.
   TIFF requires uniform full TileWidth×TileLength tiles (extra pixels are
   padding); OpenSlide/ImageScope read a sub-full-size JPEG as a "dimensional
   mismatch" and render black/garbled edges. codecTileEncoder now edge-replicates
   partial tiles up to the full tile size before encoding, across every
   engine-backed re-encode path (convert --to svs|tiff|ome-tiff|cog-wsi,
   --factor, downsample, crop, --to ife). DZI/SZI use partial edges legitimately
   via internal/dzi's own encoders and are unaffected.

2. Aperio thumbnail at IFD 1. Genuine Aperio SVS always carries the thumbnail as
   the second IFD; ImageScope classifies IFD 1 positionally as the thumbnail, so
   a source without one (BIF, IFE/Iris) left the first reduced level at IFD 1 and
   ImageScope dropped it. emitSVSThumbnailAtL0 now synthesizes a 1024px baseline-
   JPEG thumbnail from L0 (via streamCropThumbnail) when the source lacks one,
   matching the genuine layout. Sources with a thumbnail are unchanged; the
   associated-image edit rebuilds pass a nil slide (no synthesis, faithful copy).

Verified pixel-faithful against OpenSlide on a Ventana BIF→svs (no dimensional
mismatch; full slide renders clean) and IFD layout matches genuine Aperio
(CMU-1). Tests: padRGBTileReplicate unit; integration TestSVSEdgeTilesAreFullSize
(edge tile JPEGs are full tile size) and TestSVSSynthesizesThumbnailAtIFD1
(thumbnail at IFD 1, no level eaten).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>