Avoid locking on failed operations

[khaeru]

The PR adds a new context manager .utils.discard_on_error() that can be used to avoid leaving scenarios in a locked state. This can occur when exceptions are raised and JDBCBackend is in use: an unclean exit means the database is left in a state that marks the scenario as "locked" for use by a particular user.

Originally developed for iiasa/message_data#446.

How to review

• Read the diff and note that the CI checks all pass.
• Read the additions to the docs including examples and ensure they are clear.
• Optionally, try a particular code operation known to fail and leave scenario(s) in a locked state.

PR checklist

• [x] Continuous integration checks all ✅
• [x] Add or expand tests; coverage checks both ✅
• [x] Add, expand, or update documentation.
• [x] Update release notes.

[codecov[bot]]

Codecov Report

Merging #488 (246be1b) into main (4274880) will increase coverage by 0.0%. The diff coverage is 100.0%.

@@          Coverage Diff          @@
##            main    #488   +/-   ##
=====================================
  Coverage   98.5%   98.5%           
=====================================
  Files         42      42           
  Lines       4506    4557   +51     
=====================================
+ Hits        4442    4493   +51     
  Misses        64      64           

Files Changed	Coverage Δ
ixmp/core/timeseries.py	98.7% <100.0%> (+<0.1%)	:arrow_up:
ixmp/testing/__init__.py	95.9% <100.0%> (+<0.1%)	:arrow_up:
ixmp/tests/core/test_timeseries.py	100.0% <100.0%> (ø)
ixmp/tests/test_utils.py	100.0% <100.0%> (ø)
ixmp/utils/__init__.py	98.6% <100.0%> (+<0.1%)	:arrow_up:

[gidden]

I think the force push means I can't access my review to change to accept?

[gidden]

Ok, was able to do a new review - thanks @khaeru !