Fix shape for datasets of references to iterators

[h-mayorquin]

Motivation

Fix #1237

This is a draft that @rly started working on that I share here for further reference.

[codecov]

Codecov Report

Attention: Patch coverage is 37.50000% with 5 lines in your changes missing coverage. Please review.

Project coverage is 90.82%. Comparing base (ff4a0aa) to head (aab88ab).
Report is 20 commits behind head on dev.

Files with missing lines	Patch %	Lines
src/hdmf/build/objectmapper.py	37.50%	4 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##              dev    #1238      +/-   ##
==========================================
- Coverage   90.87%   90.82%   -0.05%     
==========================================
  Files          42       42              
  Lines        9524     9529       +5     
  Branches     1921     1923       +2     
==========================================
  Hits         8655     8655              
- Misses        576      580       +4     
- Partials      293      294       +1     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[h-mayorquin]

Checking the errors it seems that we were too quick to dismiss the multi-dimensional case, @rly .

[rly]

I haven't looked deeply, but I suspect that because VectorData shape allows for 1D, 2D, 3D, or 4D data (ref), and there are datasets of references in NWB such as the electrode_group column in the Units table that extend VectorData but do NOT specify that the dataset should be 1-dimensional (ref), that spec_shape is [[None], [None, None], [None, None, None], [None, None, None, None]] inherited from VectorData, and the new check is triggered.

1. We should amend the NWB schema to restrict certain table columns like Units.electrode_group to be 1-D.
2. Instead of checking the spec, we should check whether the dataset of references being written has more than 1 dimension. We should probably put that here: https://github.com/hdmf-dev/hdmf/blob/dev/src/hdmf/backends/hdf5/h5tools.py#L1252 . Note that the dataset being created has shape (len(data), ) which means we already assume they are 1-D datasets.

Related TODO items:

1. Warn if creating a spec with a dataset of references that allows more than one dimension.

I'll take a look at this in the next couple days.

[h-mayorquin]

I think #1270 already fixed the issue that this PR was aimed to solve. Fixed.