TrySaveStateAsync and TryDeleteStateAsync should only handle "rejections" based on ETag, not all RPC exceptions

[rynowak]

Expected Behavior

TrySaveStateAsync or TryDeleteStateAsync should return false iff the operation was rejected due to a concurrency violation. The design intent of these methods is to treat expected failures (etag mismatch) separately from unexpected failures (network hiccup).

Users should be able to use the true/false return value to distinguish whether state was saved, and only use try/catch if they choose to handle arbitrary exceptions.

Actual Behavior

TrySaveStateAsync swallows all RpcExceptions

TryDeleteStateAsync swallows all Exceptions

Steps to Reproduce the Problem

Issue a call to TryDeleteState that's totally invalid (say an invalid key format), or when the network is disconnected. It will be surfaced as a false return value instead of an exception.

[rynowak]

Tagging this as a breaking change (not sure why it was missing). Moving from one valid (but not ideal) to another valid behavior is a breaking change 😆