Nrt

------------------
- nrt.data now lives in its own namespace package (nrt-data). Subpackage removed,
and namespace enabling in multiple __init__.py files
- A few edits (e.g. pkg-data and MANIFEST.in) as a consequence of removing
nrt.data subpackage
- Transitioned from setup.py to pyproject.toml
- Package version now maintained in pyproject.toml
- Adapted integration fixture to new test data access method (+ band names slightly changed
B4 is now B04, etc)
- Version number now managed via git tags using setuptools-scm
- Removed all LFS hack in .readthedocs, github actions, etc. No longer needed
given that the package itself no longer contain data

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

- Various improvements to documentation and package metadata suggested by JOSS
review (new example using PC archive and STAC catalogue, Contributing guidelines,
explicit citation information, etc)
- Small documentation improvement for CCDC monitoring
- New update_mask argument and moniroting class attribute to allow monitoring to
continue regardless of boundary crossing
- Exclude version 0.59 of numba in requirements due to known bug in pararrel accelerator.
See https://github.com/numba/numba/issues/9490

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

- np.linalg.inv replaced by the more recommanded np.linalg.solve in many places
- Integration of numba parallel accelerator in most fitting functions (new argument
to control number of threads in the .fit method of BaseNrt class)
- Possibility to pass kwargs to function of data module that load xarray.Datasets
objects (particularly useful to specify chunking and get a dask based object)
- New example in gallery on parallel fitting
- Add a minimum python version requirement (>=3.9). 3.8 not longer tested
- New set of functions to generate synthetic data (single time-series and DataArrays)
added to the data module
- Gallery section added to the documentation, with one example on use of synthetic data
for nrt simulation and computation of performance metrics

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

- The report() method can now receive a list of layers as argument to select the
attributes of the monitoring instance to be retrieved and written to disk. Backward
compatibility of the .report() method is not affected but backward compatibility
of the private ._report() method, used to prepare the array is broken

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

- First pypi release