Dlisio

-------------------
* Full support for LIS Spec Block sub-type 0 and 1.
* dlisio now uses the python type 'bytes' to represent the LIS79 defined
datatype 'mask'.
* The content of DFSRs is checked more thoroughly before attempting to read the
curves.
* Integral parts of LIS Entry Block are sanity checked upon parsing.
* Fixes a bug that resulted in the curve parsing routines allocating to little
memory for the numpy array. That ultimately resulted in a segfault. This bug
affected both DLIS and LIS, but only occurs for files that are written in a
specific way.
* Fixes a bug in the DLIS curve parsing that caused dlisio to incorrectly
restore state after attempting to read a broken frame from the file,
resulting in the following frames being incorrectly written to the outputted
numpy array.

-------------------
* ``lis.load`` now raises by default when a file cannot be successfully
indexed. In previous versions only a warning was issued and a partially
indexed file was returned.
* ``lis.load`` now accepts a ``dlisio.ErrorHandler``-instance similar to
``dlis.load``. This allows the caller to decide how load behaves when a file
cannot be properly indexed.
* dlisio has learned to parse the following LIS records: Operator Command
Inputs, Operator Response Inputs, System Outputs to Operator and FLIC
Comment.
* LIS Data Format Specification Records have been given a more user-friendly
interface through ``lis.DataFormatSpec``.
* dlisio has learned to read LIS curves where the index is recorded in depth
recording mode 1.
* dlisio has learned to read LIS curves that are sampled at a higher rate than
the recorded index (Fast Channels)
* dlisio has learned to read LIS curves with non-scalar sample-values.
* dlisio has learned to read LIS curves containing strings.
* Added safe-guards against ZeroDevisionError's in lis.curves
* The predecessor and successor bits in the LIS Physical Records Headers are
checked for consistency during the indexing routines.
* Error messages emitted from the indexing routines at ``lis.load`` are now
properly communicated to the caller.

-------------------
* Added support for reading the LIS79 record types: Job Identification,
Wellsite Data and Tool String Info.
* Fixes a bug that led dlisio to attempt to read curves with multiple entries
in each sample, even though such support was never implemented. Now a
NotImplementedError is raised.

-------------------
* Solves an issue in the Windows deploy pipeline on Appveyor that resulted in
the pipeline failing to upload the python wheels to PyPi.

-------------------
* Added an initial pass at a Log Information Standard 79 (LIS79) reader. Like
the DLIS reader, the new LIS reader is mainly implemented in C/C++ with
python bindings on top. The reader is not feature complete at this point. But
it can read most curves, with a few exceptions (see the docs of
dlisio.lis.curves). Basic metadata such as LIS Header/trailer Records (RHLR,
RTLR, THLR, TTLR, FHLR, FTLR) can also be read. Support for more complex LIS
Records such Information Records is not yet added. The LIS reader
resides in the python submodule ``dlisio.lis``.
* The python module is restructured to accommodate the new LIS reader. Most
notably, all DLIS related functionality is moved to the submodule
``dlisio.dlis``. I.e. this release breaks the main entry point of dlisio
``dlisio.load`` which from this release and onwards is moved to
``dlisio.dlis.load``. For a full overview of the restructuring see
commit ``736d545``.
* The documentation on readthedocs_ has been given an overhaul to fit the new
module structure and LIS documentation is added.
* Added support for DLIS NOFORM objects.
* Better debug information for broken DLIS files.
* Better error message when passing a directory as path to ``dlisio.dlis.load``.
* Nicer error message when failing to construct datetime objects due to invalid
dates in the file.
* Added support for python 3.9
* Dropped support for python 3.5
* Restructuring the C/C++ core of dlisio, please refer to the git log for a
full overview of the restructuring.
* The C and C++ targets are merged into one target ``dlisio``, and
``dlisio-extension`` ceased to exist.

-------------------
* Fixes a bug that caused ``dlisio.load`` to fail on files >2GB on Windows.
* dlisio can now read data from truncated files, this feature is opt-in.
* dlisio can now read data from files that are padded at the end, this feature
is opt-in.
* How dlisio handles spec-violations in files is now customisable.
* ``describe()`` includes attributes units
* ``Batch`` has been renamed to ``physicalfile``.
* ``dlis`` has been renamed to ``logicalfile``
* ``dlis.match`` (now ``logicalfile.match``) is deprecated in favor of
``logicalfile.find``