[GH-1172] Bulk Ingest Workflow Outputs

[ehigham]

RR: https://broadinstitute.atlassian.net/browse/GH-1172
The sarscov2_illumina_full pipeline has many outputs, many of which are files. Ingesting them one at a time is really slow and so Jade recommend using the "bulk" insert endpoints.
In this change:

• Add bulk insert endpoint wrapper
• Gather all (unique) files in a workflow and bulk ingest them
  • the same file can appear in more than one location in pipeline outputs
  • it's an error to try to ingest the same file twice.
  • We're limited to 1000 files per bulk ingest, Muscles says it is "probably fine" to pin ingest sizes to this, so the implementation uses batches of 1000. Probably should test if there's a sweet spot but it's good enough for now.
• Add mime-type support for new file extensions
  • There are some that are not determinable from the file extension, so in the test I've added a way to exclude those
• Refactor the type dispatch logic into a function traverse for re-use.

Note:

• In this change, I'm suggesting that we give TDR bucket reader access instead of reader access for each object in the outputs. This may be controversial.

[rexwangcc]

Did one pass, looks good to me!