Changes related to sourcekey and visible columns

[RFSH]

In this PR,

• Modified the sourcekey support to allow changing entity, aggregate, and self_link at the same time (#909).
• Improved the default visible columns heuristics regarding the display key (more info here).

In #909 we talked about allowing source definitions to refer to other definitions like this

"tag:isrd.isi.edu,2019:source-definitions": {
  "sources": {
    "S_core_fact": {
      "source": [{"inbound": ["CFDE", "core_fact_id_namespace_fkey"]}, "nid"]
    },
    "S_taxonomy": {
      "markdown_name": "Taxonomy",
      "source": [
        {"sourcekey": "S_core_fact"},
        {"inbound": ["CFDE", "core_fact_ncbi_taxonomy_core_fact_fkey"]},
        {"outbound": ["CFDE", "core_fact_ncbi_taxonomy_ncbi_taxon_fkey"]},
        "nid"
      ]
    },
    "S_taxonomy_array": {
      "sourcekey": "S_taxonomy",
      "aggregate": "array_d"
    }
  }
}

What I pushed is just a simple syntactic sugar to avoid copy/paste, and so when this source is used in a visible-column, there is not signal that it's using a different sourcekey. So It cannot properly do the path sharing. But if we want this to also properly handle path sharing, we would need to think more about it and decide how it should be done.

[RFSH]

After discussing these changes, we decided to revert the extra changes I included in this PR.

While the change related to column heuristics is something we should do, we have to make sure we're doing it consistently for row-name as well. If possible, we would like to keep these two as close as possible.

And the change to allow sourcekey inside source-definitions requires more discussion.