Test concat of dimension coordinate not backed by an index

[TomNicholas]

Adds a test showing a problem in xarray - if I build two datasets that have dimension coordinates which are explicitly not backed by an index (which is possible by passing the new xr.Coordinates object explicitly to the constructor), then upon concatenating these, xarray sees the dimension coordinate and attempts to automatically rebuild a new index (which will fail in my case). I need to be able to turn off that behaviour somehow.

[TomNicholas]

The simplest test that shows this issue is described in pydata/xarray#8871

EDIT: And a proposed fix is pydata/xarray#8872

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.12%. Comparing base (5eb15bb) to head (85472bc).
Report is 6 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #44      +/-   ##
==========================================
+ Coverage   87.66%   88.12%   +0.46%     
==========================================
  Files          13       13              
  Lines         851      893      +42     
==========================================
+ Hits          746      787      +41     
- Misses        105      106       +1     

Flag	Coverage Δ
unittests	88.12% <100.00%> (+0.46%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[TomNicholas]

This PR now uses a branch on my fork of xarray, which contains the (unmerged upstream) PRs pydata/xarray#8711, pydata/xarray#8714 and pydata/xarray#8872. All three of these are needed to get the new test in this PR to pass.

[TomNicholas]

I'm going to merge this, making this package depend on my branch of xarray. Once those changes are merged upstream I can unpin it from that specific branch, see #14 (comment).