Hepdata-converter

What's Changed
* Use GitHub Actions by alisonrclarke in https://github.com/HEPData/hepdata-converter/pull/36
* YODA2 writer by 20DM in https://github.com/HEPData/hepdata-converter/pull/53

New Contributors
* 20DM made their first contribution in https://github.com/HEPData/hepdata-converter/pull/53

**Full Changelog**: https://github.com/HEPData/hepdata-converter/compare/0.2.3...0.3.0

This release improves the display of labels in the ROOT export, following the transition to Python 3. Previous versions 0.2.1 and 0.2.2 displayed labels as `b'label'`. This version removes the `b` and the surrounding quote marks. As with [version 0.2.1](https://github.com/HEPData/hepdata-converter/releases/tag/0.2.1) and [version 0.2.2](https://github.com/HEPData/hepdata-converter/releases/tag/0.2.2), it is recommended to use this release with the Docker image [hepdata/hepdata-converter:0.2.0](https://hub.docker.com/r/hepdata/hepdata-converter).

This release fixes a bug in version 0.2.1 where an exception `'TypeError: can only concatenate str (not "bytes") to str'` was raised when the `hepdata_doi` option is passed as an optional argument to the `convert` function, as done in the code for the [HEPData web app](https://github.com/HEPData/hepdata). As with [version 0.2.1](https://github.com/HEPData/hepdata-converter/releases/tag/0.2.1), it is recommended to use this release with the Docker image [hepdata/hepdata-converter:0.2.0](https://hub.docker.com/r/hepdata/hepdata-converter).

This is the first release for Python 3 and it drops support for Python 2. It is recommended to use this release with the Docker image [hepdata/hepdata-converter:0.2.0](https://hub.docker.com/r/hepdata/hepdata-converter), based on the Docker image [rootproject/root:6.22.02-ubuntu20.04](https://hub.docker.com/r/rootproject/root), which contains ROOT v6.22/02, Python v3.8.2, YODA v1.8.3, and hepdata-validator v0.2.2. However, the release can be used without Docker if these dependencies are installed independently.

This release is compatible with the new hepdata-validator v0.2.1, which adds versioning of JSON schema. By default, the hepdata-converter will validate using the latest JSON schema available in the hepdata-validator. However, an option can be passed via the CLI or the `convert` function to use an older JSON schema version. Updates have also been made for YODA v1.7.7 and PyYAML v5.3.

Support has been added for a "Custom Rivet identifier" as a qualifier of a YAML data file, which will override the default value in the YODA export. The bin number written in the `ErrorBreakdown` annotation now excludes bins with non-numeric dependent variable values. Improvements to the PyPI release have been made and several small fixes to the conversion code have also been made, such as:

* Allow for the possibility of a negative `symerror` in CSV writer.
* Allow for two independent variables with different names when transforming a square matrix in oldhepdata parser.
* Check number of y values does not exceed number of x values in YODA writer.
* Remove `$` and `\` characters from table names as filenames or directories in CSV and ROOT writers.
* Account for the case of a `symerror` given as an empty string in CSV writer.
* Allow for the possibility of Unicode characters in error values for ROOT and YODA writers.
* Improve use of newline characters in YODA output format.
* Skip defining histogram bins for non-numeric y values in ROOT writer.

* yoda writer: write uncertainty breakdown as ErrorBreakdown annotation (thanks to Louie Corpe).
* array writer: add process_error_labels() to ensure uniqueness, used by CSV/ROOT/YODA writers.
* root writer: improve code for case of unordered or non-overlapping bins.
* root writer: demand all bins have 'low' and 'high' before writing ROOT histogram.
* root/yoda writer: convert numbers given as strings to floats.
* docs: refinements to Docker installation instructions.
* oldhepdata parser: keep values and errors as strings (no float conversion) to maintain precision.
* root/yoda writer: treat one-sided errors (HEPData/hepdata122) as numerical zeros.
* root writer: use ZLIB (not LZ4) compression for better compatibility with older ROOT versions.
* docs: pin matplotlib<3.0.0 in setup.py to fix "Read the Docs" build.
* oldhepdata parser: allow one-sided errors (HEPData/hepdata122) given as '-'.
* csv writer: use default quotechar='"' and quoting=csv.QUOTE_MINIMAL.
* hepdata-converter-docker: upgrade to ROOT v6.14/06 and YODA v1.7.3.
* testdata: update all test data files to account for all converter changes above.
* global: version bump to 0.1.34.

Signed-off-by: Graeme Watt <graeme.wattdurham.ac.uk>