Cognite-sdk

Added
- All datapoints retrieve methods (except `retrieve_latest`) now support status codes. Note: Support for *inserting*
datapoints with status codes will be released shortly. There are three new arguments:
* `include_status (bool)`: Toggle the return of status code and -symbol on/off, only valid for raw datapoints.
* `ignore_bad_datapoints (bool)`: For raw datapoints: Whether to return those marked bad (or not).
For aggregates: Whether the time periods of bad datapoints should affect aggregate calculations (or not).
* `treat_uncertain_as_bad (bool)`: Toggle whether datapoints marked uncertain should be regarded as good or bad.
- The `to_pandas` method for `Datapoints`, `DatapointsList`, `DatapointsArray` and `DatapointsArrayList` now accepts
a new parameter, `include_status (bool)`, that controls whether to include status codes & -symbols as separate columns.
- New datapoints query class, `DatapointsQuery`, to make writing custom queries easier, type-safe and more robust,
as opposed to passing dictionaries (of settings).
Deprecated
- Passing *custom* datapoints queries using dictionaries is deprecated and will be removed in the next major release.
Consider refactoring already to `DatapointsQuery`. Example: `{"id": 12, "aggregates" : "min", "granularity": "6h"} ->
DatapointsQuery(id=12, aggregates="min", granularity="6h")`.

Fixed
- When using TimeSeries objects without `external_id` as part of the `variables` parameter in a synthetic datapoints
query, a `CogniteNotFoundError` would most likely be raised, due to `None` being silently cast to a string. It now
raises a friendly `ValueError`.
- An invalid expression could be created when using multiple variables in a synthetic datapoints query. This happened
while substituting the variables into the expression; this was done one at a time, leading to later replacements
possibly affecting earlier ones. Now all variables are substituted at the same time/in a single call.
Improved
- Passing sympy symbols as part of the variables mapping (in synthetic datapoints queries) is now documented properly
and "officially supported".

Fixed
- Inserting sequence data using `insert_dataframe` would by default drop all rows that contained at least one missing value.
This has now been fixed to only remove rows where all values are missing.

Fixed
- `AssetsAPI.create_hierarchy` now properly supports `AssetWrite`.

Improved
- Type validation of identifiers

Fixed
- Several methods for `DatapointsArray` that previously failed for string datapoints due to bad handling
of numpy `dtype`-to-native conversion.