Fix NN ensemble training and learning on one-document corpus

[juhoinkinen]

Training or learning NN ensemble projects on a corpus consisting of only one document has been failing due to UnboundLocalError from Keras.

On Annif side the reason for this was an error for calculating the number of batches in LMDBSequence, which was based on a wrong self._counter value: when loading the LMDB the counter value was obtained as the index of the last element, which resulted in the counter value to be too small by one (for a corpus of one document self._counter = 0, and thus len(seq) = 0). So, Keras did not find any batches to iterate over, but raised the (spurious) error.

This PR fixes the off-by-one error enabling training and - more importantly - learning nn_ensemble projects on one document only (learn is typically called on one doc at a time).

Also raises NotSupportedException for attempts of training or learning with empty corpora.

Closes #504 and #505.

[codecov]

Codecov Report

Merging #506 (2dcb990) into master (a1348aa) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #506   +/-   ##
=======================================
  Coverage   99.49%   99.49%           
=======================================
  Files          78       78           
  Lines        5687     5695    +8     
=======================================
+ Hits         5658     5666    +8     
  Misses         29       29           

Impacted Files	Coverage Δ
annif/backend/nn_ensemble.py	100.00% <100.00%> (ø)
tests/test_backend_nn_ensemble.py	100.00% <100.00%> (ø)

---

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 a1348aa...2dcb990. Read the comment docs.

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

No Coverage information
0.0% Duplication