v0.1.1

- Enable libedit for arrow-key/command-history support in the interactive
  shell (previously hardcoded off, arrow keys printed raw escape
  sequences)
- README: documented the single-EMBED() CTE pattern for fast
  similarity-search queries over a table (avoids re-running EMBED() per
  row)
- README: clarified that EMBEDDING_SIM() example numbers are
  backend/model-specific, not universal (onnx/MiniLM vs llama/nomic
  differ in absolute scale, both are correct for their own model)

v0.1.0 — SQLite CLI extension: DMPHON, EMBEDDING_SIM/DIST, EMBED()

- EMBED() with two embedding backends: ONNX Runtime (default, matches
  Ragger's pipeline) and libllama (GGUF models), selected via
  embedding_embedder config key
- Vector storage: f32/f16/bf16/int8 output types, embedding_offset,
  embedding_skip_renorm, embedding_dims (Matryoshka-style truncation)
- DMPHON() double metaphone, EMBEDDING_SIM()/EMBEDDING_DIST() cosine
  similarity/distance over stored blobs
- Built on an unmodified vendored SQLite amalgamation + shell.c
- Verified building and running on macOS (Apple Silicon) and Debian 13