v0.0.1acerebra2026-06-04

Multi-section chunking fix and config system

Fixes the critical bug where every document produced only one chunk, renames parse_warnings_json to parse_warnings, and adds a persistent config system with vault discovery, cerebra status, and a git-repo guard on init.

v0.0.1a2026-06-04

·_build_sections rewrite: clean per-boundary stack algorithm — 39 docs now produce 745 chunks (was 1 each)
·Migration003: rename documents.parse_warnings_json → parse_warnings
·4-step vault resolution: --vault > CEREBRA_VAULT > config.toml > error
·cerebra config set/get vault: persistent default in ~/.config/cerebra/config.toml
·cerebra status: sources, chunks, records, last ingest, schema version, WAL mode
·cerebra init git-repo guard: walks ancestors for .git/, --force bypass
·docs/agent/deviations/v0.0.1a.md: deviation log introduced as per-pass discipline