RFE: test client display of wells from run-less plates, consider other kinds of data

[mtbcarroll]

In investigating #10564, in which incomplete deletion of plate runs was leaving wells behind, Insight was behaving oddly and not showing the thumbnails in a neat array, but instead even drawing thumbnails on top of each other -- perhaps the server-side #10564 bug was just leaving things in too strange a state, though.

Still, if it is possible for a plate to have wells without having runs, or maybe even wells that don't have well samples, then perhaps such data should be added to the portfolio of what we test. Our current test data, that I've seen at least, has plates always having exactly one run, and wells always having samples (also exactly one?). We needn't expand to every edge case that can technically fit into the model, but I don't know if we are at least making sure that our test data is qualitatively covering what users might really throw at OMERO, even if we have to construct some of it with the fake readers or by XML hacking.