OpenTelemetry trace exporter does not reconnect to backend if connection is lost

Currently if we restart the tracing setup (tempo+grafana+prometheus), we then also have to restart all services that export traces to it.

Ideally the exporters on individual services (such as boostd, booster-http, booster-bitswap) should reconnect to the tracing backend, when it becomes available.

Start by locating the OpenTelemetry exporter setup used by boostd, booster-http, and booster-bitswap, then reproduce the issue by restarting the tempo, Grafana, and Prometheus tracing setup. Done means each service reconnects and resumes exporting traces when the tracing backend becomes available without restarting the service.

Written by the indexing model from the issue text.