Transfer Excel data I/O from message_ix

[khaeru]

This PR streamlines and improves file input/output in the MESSAGE stack:

• Scenario.read_excel(), .to_excel() are brought to ixmp from message_ix (see Transfer Excel data I/O to ixmp message_ix#321).
  • Can now be used with any ixmp Scenario.
  • The code is improved in several ways, e.g. it can now distinguish between 1-D index sets and indexed sets; degrades gracefully by giving readable log messages and exceptions.
• TimeSeries.read_file is added as a replacement for ixmp.utils.import_timeseries.
• New: commands import and export that allow performing the above operations from the command-line.
• New: documentation that clearly describes the file formats supported (and not supported) for I/O; docstrings for all touched methods.
• New: TimeSeries.add_timeseries(..., year_lim=(2020, 2080)), which brings the minimum/maximum year functionality previously only supported by import_timeseries() to non-file-based operations.
• See Try to fix failing build on AppVeyor #290 & thanks @zikolach for:
  • ixmp.jar is updated to clean up Java exception handling.
  • {,JDBC}Backend gain a method, get_log_level().

Closes #284.
Replaces #286.

How to review

• Build and read the documentation.

• Check that any I/O scripts still work.

  One possible breaking change: import_timeseries() previously automatically prepended the string 'R11_' to imported region names, unless the name was 'World'. This is specific to MESSAGEix-GLOBIOM (not message_ix), so it is removed.

PR checklist

• Tests added.
• Documentation added.
• Release notes updated.

[codecov]

Codecov Report

Merging #289 into master will not change coverage by %.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #289   +/-   ##
=======================================
  Coverage   97.38%   97.38%           
=======================================
  Files          39       39           
  Lines        3979     3979           
=======================================
  Hits         3875     3875           
  Misses        104      104           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 386f3fe...386f3fe. Read the comment docs.

[zikolach]

Didn't review code itself yet. Commented on docs about format.

[zikolach]

Thanks Paul for implementing and explaining details. Overall it looks very good and as a big step forward!

I didn't test assuming new read/write of model data is covered with tests.

In general would be good to allow splitting items data into multiple sheets as we can reach rows limit. We could avoid to have separate sheet with item name-type mapping by embedding type into sheet name, e.g. type|name|n or similar convention, where type is item type, name - item name and n is a number of sheet for "large" items.