feat: add StringSequenceToEmbedding transformer and layer

[mruiyangyou]

Summary

Adds a new StringSequenceToEmbedding transformer/layer pair that parses a delimited string of pre-computed embedding vectors into a dense (seq_len, embedding_dim) float tensor.

• Spark transformer (StringSequenceToEmbeddingTransformer) and Keras layer (StringSequenceToEmbeddingLayer) with parity behaviour.
• Input format: vectors separated by sequence_separator (default ,), floats within a vector by separator (default |). Example: "1|2|3,4|5|6" with seq_len=2, embedding_dim=3 → [[1,2,3],[4,5,6]].
• Pads short sequences with pad_value (default "0"); truncates long ones.
• Optional reverse mode reverses only the non-pad portion of each sequence (useful for chronological → recency-first ordering).
• Layer drops a trailing size-1 input axis to match the StringToStringListLayer convention, so (None, 1, 1) inputs produce (None, 1, seq_len, embedding_dim) outputs without a downstream squeeze.
• Empty tokens (from leading/trailing/repeated separators or fully empty cells) are replaced with pad_value before tf.strings.to_number to avoid StringToNumberOp failures at graph execution.
• README updated with a row in the supported layers table.

Test plan

• tests/kamae/tensorflow/layers/test_string_sequence_to_embedding.py — 9 unit tests covering default/custom separators, padding, truncation, reverse, trailing-1 squeeze behaviour, empty/malformed inputs, config round-trip, and invalid args.
• tests/kamae/spark/transformers/test_string_sequence_to_embedding.py — 6 tests including Spark/TF parity across separators, padding, and reverse modes.
• tests/kamae/tensorflow/test_layer_serialisation.py — added the new layer to the serialisation matrix; passes test_all_layers_tested_for_serialisation.
• No regressions in existing string_to_string_list tests.

🤖 Generated with Claude Code

[georyetti]

Hey @mruiyangyou - this looks good at a glance, but we are gearing up to do a big new kamae v3 release in the next few weeks or so. Can this wait till after these? I can aid migrating this to keras 3.