v6: dynamic index quantization is never parsed into VectorConfig

VectorConfig.CustomTypeAdapterFactory.read looks for the bq/pq/sq/rq keys at the top level of vectorIndexConfig. That is correct for hnsw and flat, but a dynamic index nests them one level deeper, under hnsw/flat, so the quantizer is silently dropped on read.

Verified against a live Weaviate 1.38.0 collection with a dynamic index and RQ enabled:

raw     hnsw.rq = {"bits":8,"enabled":true,"rescoreLimit":20}
typed   quantization = null

The same root cause loses Dynamic.flat's own quantizer.

Also dropped on read at the dynamic level: vectorIndexConfig.distance.

Fixing this needs a design call first: under dynamic, hnsw and flat can carry different quantizers, but VectorConfig has a single quantization() slot.

Found while working on #598, which adds getConfigAsJson as a raw-JSON escape hatch — that unblocks callers who need the complete picture, but does not fix the typed mapping.

Start at VectorConfig.CustomTypeAdapterFactory.read and compare how dynamic, hnsw, and flat configurations are represented. Resolve how the single VectorConfig.quantization() slot should handle different dynamic quantizers, then verify typed reads retain the relevant quantizer and dynamic distance; use #598's raw-JSON escape hatch as context.