Metrics has inconsistent error reporting mechanisms

[aabmass]

Some places we raise exceptions (MeterProvider.shutdown()), some places we return a bool (MeterProvider.force_flush()). These may be the only ones, but we need to dig in and determine if this is intentional.

I know trace just uses bool and exporters have a return result. I am personally not a fan, when you can just return meaningful exceptions :)

[srikanthccv]

shutdown in tracing also doesn't return anything. I think we discussed it here #2406 and agreed to raise an exception. I believe the reason force_flush returns bool is that we tried to be consistent what exists already for tracing/logs. Since force_flush could be called multiple times during the application lifetime, just logging exception seemed fine at the time.

[ocelotl]

I think for consistency with tracing we should keep the bool in force flush.

[aabmass]

Thanks @srikanthccv. Discussed with @ocelotl offline as well and agreed. We'll continue to return bool but allow exceptions to bubble up since they are far more descriptive than true or false