Dyntastic

- Added support for `__table_name__` being a `Callable[[], str]` to allow
dynamic table name
- Added support for batch reads and writes
- Fixed `consistent_read` behavior for `safe_get` (previously was always set to
`True`)

- Added support for multiple subclasses within one table (`get_model` function)

- Fixed serialization of dynamo types when using Pydantic aliases

- Fixed compatibility with Pydantic aliases

- Added support for nested attribute conditions and update expressions
- Fixed bug where `refresh()` would cause nested Pydantic models to be
converted to dictionaries instead of loaded into their models
- Added Pydantic aliases (models will all be dumped using pydantic's
`by_alias=True` flag).

**BREAKING**: Accessing attributes after calling `update(..., refresh=False)`
will trigger a ValueError. Read below for more information.

- Added built in safety for unrefreshed instances after an update. Any
attribute accesses on an instance that was updated with `refresh=False`
will raise a ValueError. This can be fixed by calling `refresh()` to get
the most up-to-date data of the item, or by calling `ignore_unrefreshed()`
to explicitly opt-in to using stale data.