Better observability of performance and errors in Edge API

[matthewelwell]

[matthewelwell]

We had to remove Sentry from the Edge API stack in production, primarily because of the outage that they had shortly after we released it into the Edge API, but since then, we have also determined that it did noticeably increase the latency of the edge API for certain customers.

We could perhaps look into their relay product as a means to reduce latency / risks associated with their stability but we would need to consider how we would architect this for each of the Edge regions.

Initially, I think we should spend some time improving the monitoring / alerting in Cloudwatch to see if that would provide sufficient information.