The current default RetryStrategy in async_rpc is to not retry. This is potentially problematic as it makes it easy to write new code that isn't resilient to transient errors or load shedding. We should consider updating this default to be DefaultRetryStrategy, which retries the typical set of retryable errors.

As part of this, we'll need to audit existing callers that use the default to ensure that retrying is acceptable in those cases, though it usually should be.