feat: BatchTransformerBase for fixed-size batching (#89)

[Chris-Wolfgang]

Adds BatchTransformerBase<TSource, TProgress> for fixed-size batch accumulation. Closes #89.

Stacked on #248 (base feature/async-disposable, the tip of the 0.14.0 base-class chain) so its regenerated baseline includes the full release surface.

What

BatchTransformerBase<TSource, TProgress> : TransformerBase<TSource, IReadOnlyList<TSource>, TProgress> accumulates source items into batches and yields each as IReadOnlyList<TSource>, flushing the trailing partial batch when the source completes — the bug everyone re-introduces when hand-rolling bulk ops.

var batched = batcher.TransformAsync(extractor.ExtractAsync(token), token); // IAsyncEnumerable<IReadOnlyList<OrderRecord>>
await loader.LoadAsync(batched, token);                                     // one bulk insert per batch

• BatchSize — configurable, default 100, validated >= 1.
• Sealed override of TransformWorkerAsync so the batching contract can't drift; derived classes only implement CreateProgressReport.
• CurrentItemCount counts batches yielded (the downstream unit of work), not input items. Cancellation observed while enumerating.

Per the issue, this likely moves to a future Wolfgang.Etl.Transformers package; it lives in Abstractions until that exists.

Verified

MINOR (additive). PublicAPI.Shipped.txt regenerated. Full Release build clean across all TFMs (0 warnings); 246 tests pass (7 new: evenly-divisible, remainder-flush, single-partial, empty source, batch counting, BatchSize default + validation).

Part of: #154

[Chris-Wolfgang]

Closing — superseded by ChunkTransformer<T> in ETL-Transformers, which already does fixed-size batching (yields chunks, flushes the partial final chunk, validates size >= 1, aimed at bulk loading). BatchTransformerBase is concrete transformer behavior and belongs in the transformers library, not in Abstractions (which is contracts + scaffolding only). The two minor deltas (IReadOnlyList return, progress support) are being raised as small PRs against ETL-Transformers instead.