Tracing startup should retry on error

[cohix]

May 05 21:54:40 data-stg-da-0 constd[1168]: Error: atmo.New: setupTracing(atmo, reporter_uri, 0.040000): collectorExporterOptions: grpc.DialContext: context deadline exceeded
May 05 21:54:40 data-stg-da-0 constd[1168]: Usage:
May 05 21:54:40 data-stg-da-0 constd[1168]:   atmo [bundle-path] [flags]
May 05 21:54:40 data-stg-da-0 constd[1168]: Flags:
May 05 21:54:40 data-stg-da-0 constd[1168]:   -h, --help      help for atmo
May 05 21:54:40 data-stg-da-0 constd[1168]:   -v, --version   version for atmo
May 05 21:54:40 data-stg-da-0 constd[1168]:       --wait      if passed, Atmo will wait until a bundle becomes available on disk, checking once per second
May 05 21:54:41 data-stg-da-0 constd[1074]: {"log_message":"(E) failed to getReport for 8080: failed to Do metrics request: Get \"http://localhost:8080/meta/metrics\": dial tcp [::1]:8080: connect: connection refused","timestamp":"2022-05-05T21:54:41.169266543Z","level":1}

[cohix]

When that error occurs, we should instead retry on a backoff a reasonable number of times before dying.