Dtscalibration

---

Added

* support for Python 3.12.
* AP sensing .tra support, as the reference temperature sensor data by this device in only logged in .tra and not in the .xml log files.
added functions in io/apsensing.py to read .tra files if they are in the same directory as the .xml files.
* more test data from AP sensing device N4386B, which do also contain their .tra log files

Fixed

* device ID bug for APSensing. Device ID is N4386B instead of C320. C320 was an arbitrary name given for the wellbore by the user.

Changed

* the `verify_timedeltas` keyword argument is now optional when merging two single ended datasets.
* removed `statsmodels` as a dependency. It is now optional, and only used for testing the `wls_sparse` solver

---

Fixed

* Improved error message when some variables cannot be shifted by the suggest_shift_double_ended function.

Changed:

* [dev] Moved to Ruff instead of isort for import sorting

---

Fixed

* Bumpversion configuration was incorrect since v3.0.0. Therefore the 3.0.1 release failed. This is a re-release of 3.0.1:

---

Fixed

* Oryx v4 double ended could have the backward measurements incorrectly cut off during loading.

---

Added

* Calibration functions are now accessed via the .dts accessor in favor of the DataStore class.
* Improved the functionality of `merge_double_ended`, by adding a check that handles measurements missing in one channel while present in the other ([171](https://github.com/dtscalibration/python-dts-calibration/pull/171))
* Support for Python 3.11
* `matching_sections` is now similarly implemented as `sections`.

Fixed

* Single-ended measurements with `fix_alpha` failed due to a bug introduced in v2.0.0 ([173](https://github.com/dtscalibration/python-dts-calibration/pull/173)).
* Headers in example notebooks and their appearance in the docs are now at correct levels
* Big clean up of the documentation. The documentation is now more consistent and easier to read.

Changed

* Notebooks now reflect the new API with the .dts accessor.
* Calibration functions now return only calibration results instead of adding the results to the Dataset inplace.
* Standardized parameter names. Reduced the freedom in choosing parameter names and dimension names in favor of simplifying the code.
* Requiring netcdf4 >= 1.6.4
* Optional dependencies of xarray that improve performance are now required by default.
* variance_stokes_linear() does not default through zero anymore.
* Refactored calibration_single_ended and calibration_double_ended()
* Moved ParameterIndices classes to calibration_utils.py
* Flox included in requirements to speed up resampling via xarray ([Xarray 5734](https://github.com/pydata/xarray/pull/5734)).

Removed

* Removed the `DataStore` module in favor of the `dts` accessor.
* Removed ds.resample_datastore() in favor of using xarray's resample function. See example notebook 2.
* Removed support for Python 3.8

Developer changes

* Added mypy to dev dependencies and CI.
* Using ruff for most of the formattting. Black is used for the notebooks. Isort is used for sorting the imports.
* Changed all import statements to be absolute instead of relative.
* Cleaned up old 'pylint: disable' comments

------------------

* Faster approach on error propagation. As alternative to computational intensive Monte Carlo runs.
* Refractored some of the indexing logic of the calibration routines.
* Rewritten single-ended and double-ended example notebooks.
* Removed support for OLS calibration, as with minor extra efforts WLS could be used, which has significant benefits.