Ecl2df

- Improve precision of datetime index using TIMESTEP (412)
- Stricten parsing of malformed parameter csv's with pandas>=1.4 (414)
- Wellcompletiondata metadata (409)
- Replace ert_shared with ert.shared following upstream change in ert (413)
- Removed python 3.6 and 3.7, added 3.10 (410)
- New module wellcompletiondata (404)
- Set Sphinx language to en (405)
- Complump defaults set to 0 (403)

Bugfix for on-premise integration testing

Bugfixes
* Handle defaulted coordinates in WELOPEN (item 3-5)
* Support two-phase dry gas in fipreports

**Full Changelog**: https://github.com/equinor/ecl2df/compare/v0.15.0...v0.16.0

New features
* Support writing smry and grid data to arrow format
* New submodule, wellconnstatus, for calculating states through time for invidividual connections from summary data
* Log messages will be split between stdout and stderr for the command line tool based on log severity.

Bugfixes
* WELOPEN parsing has been aligned with Eclipse's behaviour.
* Duplicated summary columns are detected and handled.
* summary will no longer raise an exception when no columns are chosen
* Fix 1OVERBO in pillars.

**Full Changelog**: https://github.com/equinor/ecl2df/compare/v0.14.0...v0.15.0

Improvements:
* [satfunc] Defaulted values in parsed saturation tables will be linearly interpolated.
* [compdat] Added support for WLIST
* [compdat] Added support for COMPLUMP
* [compdat] Added support for wellname wildcars in WELOPEN
* [gruptree] Add support for BRANPROP and NODEPROP
* All modules will now export empty dataframes to a CSV file when relevant (previously only some did)

Bugfixes:
* [rft] Fixed export of extra csv files in debug mode
* [grid] datetime.datetime did not work for API users. Fixed.
* [summary] Fixed problem with dates beyond year 2262 and end-dates not on frequency boundary.