Dataset upload fix

[ryanrdoherty]

This PR addresses two issues:

1. Dataset upload takes too long because we are allocating only one dataset_value_id at a time from the Oracle sequence. Fix is to allocate a batch of IDs at a time. How many depends on an estimate of the size of the dataset (or exact match for basket- and step-sourced datasets).
2. AnswerValue has long had two methods that load all IDs of the answer value into memory at one time as a List. This is problematic for memory and bad practice. One of the methods could be removed (string processing moved elsewhere); the other was changed to return a closable Iterator<String[]>, which callers can read one PK at a time off of, or concatenate together if they want to live dangerously.

Note this does NOT include the change where basket and strategy results are transferred directly to datasets. Instead they are written to intermediate files (which baskets were doing before), and parsed back in by ListDatasetParser. Thus the contents are still being written into the CLOB field in the datasets DB table.

[ryanrdoherty]

Note this PR helps solve VEuPathDB/ApiCommonModel#13 but not completely. We really need that SQL fix in the model XML to resolve the query performance.