Libhxl

Major changes:
- Update hxlappend to take a list from a file
- Fix hxlsort date handling
- Fix column handling in hxlappend
- Improve Google Drive URL handling
- Avoid "None" values during a merge
- Change Row.get() to return first _non-empty_ match
- Move HXLTagsNotFound exception so that we can use it in a UI flow
- Preserve lineends inside a value in CSV file (substantial rewrite of I/O handling)

Bumped version to reflect significant changes in recent updates.
- now use `hxl.data(url)` instead of `hxl.hxl(url)` to create a new dataset
- some filters and filter methods renamed
- filters substantially refactored internally to use more common infrastructure (and eliminate bugs)
- more doc comments

- Multiple bug fixes.
- All tests pass now in Python2 and Python3.
- Python3 is now the default environment.
- Sorting a mixture of strings and numbers now works.

Improved output from the hxlvalidate command to allow different levels of validation.

Added the hxldedup command and associated dedup method.

Many bug fixes.

Fixed bug in replacement table handling.

Major new feature: the hxlreplace script has changed its arguments, and can now take a mapping table to do multiple replacements in a single pass.