default result type to :none

[mitchellhenke]

elixir-lang/elixir#9028 shows a Sentry bug from #174 in which the current behavior causes messages to build up in the calling process. This is likely primarily affecting processes calling Sentry.capture_exception/1, as Sentry.LoggerBackend will handle the messages on handle_info/2 and Sentry.Plug will generally live in a process that's about to exit.

The default should have been :none all along, so this PR fixes that, and the documentation already discusses the potential issues with :async and Task.Supervisor.async_nolink/2

This closes #360

[mitchellhenke]

Note to myself, we should consider returning the generated ID of the event regardless of the result type to make stuff like #365 easier

[jeroenvisser101]

@mitchellhenke would love to be able to configure defaults like these. Was looking for a way to change this, as we tried reporting unknown messages to Sentry (which ended up being an infinite loop). Being able to configure this allows it to not require a major release (yet), as the default could stay the same. Either way, we would also prefer to use :none, so configurable or not, we would love to see this in a future version :)

[mitchellhenke]

@jeroenvisser101 a configuration for that is possible, but the default needs to change either way due to the inherent issues mentioned in the issue 🙂