Libhxl

* improve support for private datasets: fix custom HTTP headers with CKAN URL munging; add HXL_HTTP_HEADER and --http-header support for all scripts
* add hxl.io.HXLIOException as base class for all HXL I/O-related exceptions, with an optional url property
* add hxl.io.HXLAuthorizationException, and throw it for private CKAN datasets or any 403 Forbidden response
* better error reporting for external validation errors

Validation:
* better error reporting when an external resource (e.g. a taxonomy to check against) is unavailable
* fix a DIV0 bug when looking for numeric outliers
* efficiency improvements

Row formulas:
* fix edge case in date parsing (month but no year)
* return "NaN" when there's an attempt to divide by zero
* stability improvements

General:
* fix a conflict with the jsonpath_rw package
* fix bug that caused an error when reading an empty dataset

Interim release:
- fix bugs in formula handling
- fix bug in append-datasets filters
- add round() function to row formulas

Bugfix release because the hxl.formulas package wasn't being included in the PyPi distro

* **New feature!** add spreadsheet-style formulas calculated from a row's contents: https://github.com/HXLStandard/hxl-proxy/wiki/Row-formulas
* update add-column filter to support formula substitutions
* improve error reporting for private HDX resources
* fix bug with is min/is max aggregator and mixed datatypes
* fix date normalisation to preserve original value (and log an error) if parsing fails
* bugfix for validation error when no row and/or column present

* added concat() aggregator for CountFilter
* added data_hash and columns_hash properties for hxl.model.Dataset, together with hxlhash command-line script
* fixed parsing of YYYY-MM-DD hh:mm:ss SQL dates
* support opening zipped CSV files (thanks to orest-d )