Libhxl

- raise a HXLHTMLException (subclass of HXLIOException) when HTML markup found
- clean dates that are integers, assuming seconds or days since epoch (with cutouts for years and month days)
- bypass RequestResponseIOWrapper, because it's causing grief; need to make sure content still gets uncompressed
- fix input options for secondary datasets in scripts (e.g. an appended or joined dataset)
- more debug logging about HXL hashtag detection
- add a new logger to hxl.REMOTE_ACCESS specifically for external URL access
- various command-line script updates and fixes

- use loglevel CRITICAL only for infrastructure failures
- don't put tracebacks in logging (avoid logger.exception())

- raise exception for HTML input (prevents tagger exploit in HXL Proxy)
- make hxl.input.munge_url() public
- remove default tags for hxlcount script so that it can just count lines of data

- support optionally filling merged areas (XLS and XLSX only)
- support optionally scanning CKAN datasets for the first HXLated resource (in hxl.data(), not make_input())
- support Google Sheets interactive view URLs for XLSX files (but not tab identifiers)
- make Excel workbook handling more efficient
- add --encoding option to all command-line scripts (tested with CSV)
- add --expand-merged option to all comand-line scripts
- add --scan-ckan-resources option to all command-line scripts
- add hxl.input.InputOption object to hold all input options
- add info method to input objects to get general info about an XLSX file (including HXLated and merged areas in each sheet)
- refactor input and model classes
- upgrade to xlrd3 version 1.1.0

- remove discontinued "encoding" keyword for json.load (Python 3.9)
- rename the hxl.io module to hxl.input (to avoid problems in Python on Windows)

- fix bug that prevented tagger from adding tags if there was a blank row before the text headers