Cognite-sdk

Changed
- Datapoint insertion changed to be less memory intensive.

Fixed
- Fixed a bug where add service account to group expected items in response.
- Jupyter notebook output and non-camel cased to_pandas uses nullable int fields instead of float for relevant fields.

Added
- Support for the error field for synthetic time series query in the experimental client.
- Support for retrieving data from multiple sequences at once.

Fixed
- Fixed a bug where datapoints `retrieve` could return less than limit even if there were more datapoints.
- Fixed an issue where `insert_dataframe` would give an error with older pandas versions.

Added
- Support for `ignore_unknown_ids` on time series `retrieve_multiple`, `delete` and datapoints `retrieve` and `latest` and related endpoints.
- Support for asset subtree filters on files, sequences, and time series.
- Support for parent external id filters on assets.
- Synthetic datapoints retrieve has additional functions including variable replacement and sympy support.

Changed
- Synthetic datapoints now return errors in the `.error` field, in the jupyter output, and optionally in pandas dataframes if `include_errors` is set.

Added
- Support for synthetic time series queries in the experimental client.
- parent external id filter added for assets.

Fixed
- startTime in event dataframes is now a nullable int dtype, consistent with endTime.

Fixed
- Fixed notebook output for Asset, Datapoint and Raw.