------
- Switched from `dtype_str` to `str(dtype)` (Pandas 0.25+ compatibility)
- Implemented `collection.items` and `collection.snapshots` as `property` to reduce initialization overhead
- `collection.items` and `collection.snapshots` are now of type `set()`
- Option to specify both `npartitions` and `chunksize` in `collection.append()`