Add stress tests

[gaurav]

We sometimes run into issues that show up only when NodeNorm is being strained (e.g. #276, #175). We should write some tests designed to stress NodeNorm out. Ideally this could be added to the integration tests in Babel Validation (https://github.com/TranslatorSRI/babel-validation), but if not we should at least add them to Yaphet's Predator instance.

One of the issues here is that Redis tends to cache requests, so we can't use anything from the existing Babel test results that are part of the integration tests, nor can we use a fixed list that we can test repetitively. Given that we're just trying to evaluate NodeNorm's networking and database lookup capabilities, it might be enough to create a test that generates a large number of random CURIEs (randomly chosen prefix from https://nodenormalization-sri.renci.org/1.5/get_curie_prefixes + random number with a mean somewhere in the 100s but a long tail into the 10,000s).

[gaurav]

• Going to Predator with random chunks of the identifier list.
• Add logging to NodeNorm to include minimum "Took [x] seconds to normalization [y] identifiers: [CURIEs]".