GAF/Source validation processing should occur with only one pass

[dougli1sqrd]

Currently the Makefile and ontobio validate are structured very group centric. So when processing say fb we do all processing that that requires to completion. This includes merging any "mix-in" datasets, for examaple: PAINT. So in the course of validating fb we also download and validate paint_fb.

In normal pipeline mode though, we also separately process paint, including paint_fb. Since we validate paint_fb above in the course of validating fb we are processing paint_fb and potentially any other mix-in source twice.

On its own processing twice is a little lame, but has been okay for quite some time. But as we have expanded features of ontobio validate and the pipeline there's been difficulties. In particular, #1253 is ultimately caused by this "double processing" issue outlined above.

The pipeline, Makefile, and ontobio should be structured so that main validation of sources only occurs once per dataset. Merging of mix-ins into main sources as output products can come as a separate step. @kltm and I will expand on solutions here to do this.

[dougli1sqrd]

It occurs to me also that doing only one pass would eliminate completely any mid-validate GAF downloads. Recall that we have had an issue where mix-in GAFs are still downloaded within ontobio.

[kltm]

@dougli1sqrd Good point--we've been burned by that before as well.

[kltm]

This will also add clarity as we add more internal and upstream sources.

[kltm]

Talking to @dougli1sqrd earlier, one idea would be to process all incoming files without any thought of mixins, get the files, then perform the reassembly as a discrete step afterwards. This would make it easier to trace issues, view intermediate products, and add new sources/products/mixins in the future.

[kltm]

Note that we need this to get to geneontology/pipeline#27

[dougli1sqrd]

In ontobio, the order of operations will make this difficult:
Currently ontobio operates in this order:

1. Produce pristine GAF,
2. Make GPI,
3. Mixin datasets (example: paint_fb.gaf merges into fb.gaf),
4. make the rest of our products (gpad, ttl)

Step 3 is what this issue addresses. But if step 4 is dependent on step 3, we will need to resolve this difficulty in order to complete this issue.