Fix image by step plots breaking on nonexistent steps

[julieg18]

Spotted this possible bug when working on #4476. See inline comment for details.

Demos were made in an environment where the experiment run had less epochs than the workspace initially. I trimmed the videos towards the end.

Main

trimmed-main.mov

PR

trimmed.pr.mov

[julieg18]

Running experiments start off with the workspace's images than go on to hold their own created images (Is that normal or is there a bug with plot diff?). This means that an revision could start off with 20 images in the beginning of an experiment run if the workspace originally had 20 than less than 20 by the end.

A user could set a multi image slider to a step that doesn't actually exist within the experiment during the beginning of the experiment run. This would cause the plots to crash once the images get updated to a smaller length. This change handles this scenario and stops plots from crashing.

[codeclimate]

Code Climate has analyzed commit e420d69 and detected 0 issues on this pull request.

The test coverage on the diff in this pull request is 100.0% (85% is the threshold).

This pull request will bring the total coverage in the repository to 95.2% (0.0% change).

View more on Code Climate.