The controller doesn't support configuring ndjson as the source of data for the pipeline

[rcrichton]

Adding support for configuring ndjson as an input gives the pipeline flexibility as the alternative methods of FHIR Search API, Bulk Data export and JDBC all expect and enforce particular constraints on the source system. NDJSON support is much more neutral as any system can easily generate this format and if it could be configured as an input in the controller this will allow a more general mechanism to send data down the pipeline from any source system.