downloads of ontology in csv format results in error if new submission is created but not yet processed

[alexskr]

When a new submission is created .csv.gz file is removed from the previous submission before the latest submission gets parsed and becomes ready. Downloading ontology in csv format will results in 500 error until latest submission is successfully parsed.
see ncbo/ontologies_api#123

This problem is usually not noticeable because latest submission typically gets parsed within a few minutes after it is created. However, in some edge cases this becomes a problem:

1. adding an a new UMLS release can take multiple days so downloads of csv is broken at that time.
2. if latest ontology fails to parse then downloads of csv files will result in error even though it is possible to browse ontology classes in the UI

possible solutions:

1. delay removing artifacts for previous submissions until after the latest submission is successfully parsed
2. at the bare minimum api should respond with 404 status instead of 500.

[jvendetti]

The generation of CSV representations of ontologies is interleaved with indexing. If an ontology fails to index, the CSV will not be generated.