Make output data source name more relevant to the spec name.

[kanitw]

This will make debugging much easier. Right now we have no idea if data_0, data_1, etc are for which data sources.

Given we're merging data source sometimes, we need to come up with a good merging rule too.
(A dumb one could be concatenating previous name.)

[kanitw]

• We should try to see if we can make data source name depends on spec name.

• Also, we can make default unit spec names includes its mark type too/

Basically, modify the first line of UnitModel constructor:

-super(spec, parent, parentGivenName, config, repeater, undefined);
+super(spec, parent, parentGivenName + '_' + (isMarkDef(spec.mark) ? spec.mark.type : spec.mark), config, repeater, undefined);

[domoritz]

• 2) Name sources by the dataset
• 1) Use explicit spec name
• (Let's not do for now) 3) Name data sources by the implicit unit spec name if it's a leaf (a node can be a leaf iff it is a unit spec) -- this name should be based on the mark like kanitw suggested above.