Improve eval

[jlewi]

Fix a number of issues with eval.

1. Measure the time generation takes

   • Latency has a huge impact on UX
   • Latency can be affected by a number of things (e.g. amount of context) so we want to be able to measure that so we can see the impact

2. Flush the logs (Fix Evaluator final error messages aren't sent to GCP - problem flushing logs? #295) in app.Shutdown because the experiment seems to be failing to push final errors to GCP

3. Don't terminate if we fail to wait for a blocklog;

   • Keep track of this inside the EvalResult proto
   • Aborting on waitForBlockLog is just causing us to need to run multiple times in order to successfully complete an experiment

4. Improve logging about progress of experiment

   • Log example index and number of examples whenever we process an example

[netlify]

✅ Deploy Preview for foyle canceled.

Name	Link
🔨 Latest commit	056cde1
🔍 Latest deploy log	https://app.netlify.com/sites/foyle/deploys/670d68f3b5ed5700082921d2