adjust check_samples function

[moritzhauschulz]

Description

This should fix the off by one error raised in #2357. Check the issue for the bug description.

With this bug fix, the following setting should yield single sample training.

  start_date: 1979-01-01T00:00
  end_date: 1979-01-01T12:00

forecast :
    time_step: 06:00:00
    num_steps: 1
    offset: 0
    policy: "fixed"

However, no checks for other edge cases involving check_samples were done. Please review with respect to potential edge cases where the fix breaks the code.

Potential reviewers: @clessig @shmh40 @enssow @grassesi

Closes #2357

Checklist before asking for review

• I have performed a self-review of my code
• My changes comply with basic sanity checks:
  • I have fixed formatting issues with ./scripts/actions.sh lint
  • I have run unit tests with ./scripts/actions.sh unit-test
  • I have documented my code and I have updated the docstrings.
  • I have added unit tests, if relevant
• I have tried my changes with data and code:
  • I have run the integration tests with ./scripts/actions.sh integration-test
  • (bigger changes) I have run a full training and I have written in the comment the run_id(s): launch-slurm.py --time 60
  • (bigger changes and experiments) I have shared a hegdedoc in the github issue with all the configurations and runs for this experiments
• I have informed and aligned with people impacted by my change:
  • for config changes: the MatterMost channels and/or a design doc
  • for changes of dependencies: the MatterMost software development channel