v0.58.0

Extract Processors from Keeper (Phase 1 Hybrid Processing)

Decouples processing logic from store I/O to enable local-or-remote execution path for summarization and OCR.

New:

• keep/processors.py — pure functions (process_summarize, process_ocr, ocr_image, ocr_pdf) returning ProcessorResult
• Keeper.apply_result() — single entry point for writing processor results to both stores
• 14 new unit tests for processor functions

Docs:

• Architecture: updated pending queue description, OCR pipeline flow diagram
• Reference: added --retry flag, OCR/analyze task types
• Quickstart: OCR provider rows (Ollama glm-ocr, MLX mlx-vlm)
• Output: JSON format example alongside YAML
• Tagging: speech-act assertion/assessment examples

637 tests passing.