Dataflow debugger

[tuner]

If no mark instances are shown in the visualization, it's a bit difficult to figure out what's the reason. The main culprit is that when an undefined field is accessed, it's interpreted as a null value, which is currently not an error. This could be mitigated by tracking the columns that the data objects should have and raise an error when an accessor tries to access a nonexistent column. However, there are also other reasons why no data is shown.

Anyway, some sort of tool to visualize the data flow and inspect random data objects (like the first/last) would make debugging much easier. Vega editor has a really fancy one, but something much simpler would be enough as GenomeSpy's data flow model is much simpler.