Improve logging on ITRB

[gaurav]

NodeNorm only appears to produce logs for 200 requests to /docs and for 404 requests on ITRB -- other requests appear to go unlogged. This might be a deliberate strategy to avoid overwhelming the logs (we still see normalization requests on Dev), but it might be useful if we could turn this on for ITRB CI at least so it's easier to access request times and other such information.

Also, those /docs requests are FREQUENT, so from a logging perspective it would be helpful to turn them off probably.

• Log identifiers that could not be normalized so we can collect information about them (/ht @kevinschaper)

[gaurav]

We now have code for logging requests and a Jupyter Notebook for analyzing the results. The biggest roadblock now is that AWS only lets us download 10K log lines at a time, which isn't enough to capture more than a few hours if Translator is being heavily used. Some options:

1. Write a script for automatically downloading logs for a period of time (by downloading the first 10K logs, then requesting the next 10K logs, and so on).
2. Come up with some external place for the logs to be stored (e.g. in OTEL, in a database started with NodeNorm, in a dedicated logging source, etc.)
3. ???