Hepdata-converter

* Same as 0.1.32 Release but edited .travis.yml and bumped version in attempt to fix automatic PyPI upload.

* oldhepdata: fix a bug calculating npts if a matrix with no x-axis.
* oldhepdata: catch an error if *qual line without values.
* oldhepdata: print line if error doesn't match expected pattern.
* oldhepdata: check for presence of *xheader and *yheader before *data line.
* oldhepdata: demand whitespace after value if asymmetric first error.
* oldhepdata: use deepcopy to avoid YAML anchors/references when transforming a square matrix (thanks to Tim Adye).
* yaml parser: allow for a bug in PyYAML where numbers like 1e+04 are parsed as strings not as floats (see https://stackoverflow.com/a/30462009 ).
* yaml parser: allow for empty YAML documents when defining table index.
* writers: support table names containing "/" by replacing by "-" in filenames.
* writers: allow for unicode characters in table headers.
* root writer: fix bug if different error labels in different rows.
* root writer: allow for numerical rounding when checking that errors are symmetric before writing a TGraph2DErrors object.
* testdata: update for changes in format made in YODA v1.7.0.
* docs: tidy up and improve documentation produced using Sphinx.
* global: version bump to 0.1.32.

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

* yaml: always use SafeLoader and SafeDumper or preferably LibYAML classes.
* yaml parser: check for documents with 'data_file' in submission.yaml:
* Skip empty YAML documents, e.g. terminating "---" line.
* Information about whole submission is optional as first YAML document.

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

* root writer: use get_hist_classes method instead of _hist_classes attribute
* root writer: don't write a graph object if no numerical data points
* oldhepdata parser: print line if mismatch in data entry elements
* oldhepdata parser: remove 'stat' label from single errors and replace keyword 'energies'
* yaml writer: write hepdata_doi and table_doi without !!python/unicode
* oldhepdata parser: allow absence of 'location', 'description', or record_ids

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

- yoda writer: write Scatter1D object if zero independent variables
- root and yoda writers: improve treatment of non-numeric data
- writers: add options to pass hepdata_doi and rivet_analysis_name

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

- oldhepdata parser: improve extraction of values of the energies keyword
- oldhepdata parser: reformat correlation/covariance matrices