Make Cook Book doc on how to use map types and how to scatter them

[ekiernan]

We do not have a good example pipeline that uses map types. It would be nice to add a how-to blurb for this data type to the Cook Book section of the WDL docs.

[kayleemathews]

Example workflow from Daniel Park on openwdl slack

The workflow:

• starts with a simple scatter on an Array of accessions
• performs (in a custom task, since WDL has no native way to do this) a group-by operation on that Array to bunch them up by accession number (since each accession might otherwise have multiple other things associated with it)
• then does a scatter on a zip of the group-by outputs (so now you've got your Pairs)
• inside the scatter it's doing value lookups on Maps
• demonstrates a common pattern I have where a WDL scatter produces an Array[String] per item, which then becomes an Array[Array[String]] which is written out as a single TSV file. adding a header to such a TSV file is a little clunky so it's handled with a concatenation task.

[kayleemathews]

There also might be an example in the Cromwell repo