index-keyspace-evolution

# Index Keyspace Evolution Use this when adding a new Fjall keyspace, extending an existing key encoding, or changing the value shape of an indexed record. ## Decide if you need a schema bump Bumping `INDEX_SCHEMA_VER` triggers auto-wipe of `.basemind/views/<view>/index.fjall/` on the user's next scan. The next scan rebuilds from the msgpack blobs — no data loss, but a one-time scan cost. Bump when: - The byte layout of an existing key changes. - A `SymbolKind` ordinal would shift (you reordered the enum — don't). - A value's msgpack shape lost a field or renamed one. Do **not** bump when: - Adding a brand-new partition (no existing entries to migrate). - Adding a `#[serde(default)]`-defaulted field to an indexed value. - Adding a new `SymbolKind` variant at the tail (the `u8` mapping is stable). ## Steps 1. **Encoder + decoder + tests in `src/index/keys.rs`** - Add `<partition>_key(...) -> Vec<u8>` and `<partition>_prefix(...) -> Vec<u8>`. - Add `parse_<partition>_key(&[u8]) -> Result<(...), …>`. - Add unit tests for round-trip + prefix isolation: insert two keys with a shared prefix that differ only in the suffix, scan the prefix, confirm the wrong sibling is excluded. 2. **Open the partition in `src/index/mod.rs`** - Add the field to `IndexDb`. - Open via `db.keyspace("<name>", PartitionCreateOptions::default())` next to the others. - If used only for reads, mark `#[allow(dead_code)]` until a writer / reader exists. 3. **Writer in `src/index/wri...