Cognite-sdk

Changed
- Error 500's are no longer retried by default, only HTTP 429, 502, 503, 504 are.
- Optimized HTTP calls by caching user agent.
- Relationship filtering is now integrated into `list` instead of `search`.
- Sequences `insert_dataframe` parameter `external_id_headers` documentation updated.
- Type hints for several objects formerly `Dict[str, Any]` improved along with introducing matching dict derived classes.

Fixed
- `source_created_time` and `source_modified_time` on files now displayed as time fields.
- Fixed pagination for `include_outside_points` and other edge cases in datapoints.
- Fixed a bug where `insert_dataframe` with strings caused a numpy error.

Added
- Relationships can now have sequences as source or target.

Changed
- Datapoints insert dataframe function will check for infinity values.
- Allow for multiple calls to .add / .remove in object updates such as metadata, without later calls overwriting former.
- List time series now ignores the include_metadata parameter.

Added
- Advanced list endpoint is used for listing time series, adding several new filters and partitions.

Added
- Datapoints objects now store is_string, is_step and unit to allow for better interpretation of the data.
- Sorting when listing events
- Added a search function in the relationships API.

Changed
- `list` and `__call__` methods for files now support list parameters for `root_ids`, `root_external_ids`.
- retrieve_dataframe with `complete` using Datapoints fields instead of retrieving time series metadata.

Fixed
- Fixed chunking logic in list_generator to always return last partial chunk.
- Fixed an error on missing target/source in relationships.

Fixed
- Fixed support for totalVariation aggregate completion.
- Changed conversion of raw RowList to pandas DataFrame to handle missing values (in columns) across the rows. This also fixes the bug where one-off values would be distributed to all rows in the DataFrame (unknown bug).

Changed
- Sequences officially released and no longer considered experimental.
- Sequences data insert no longer takes a default value for columns.

Fixed
- Tokens are sent with the correct "Authorization" header instead of "Authentication".