Consider removing the periodic retries in runtime client SubmitTx

[ptrus]

Ever since starting to use hash message payloads as libp2p msg ids (#3280), the runtime client re-sending messages is ineffective:

oasis-core/go/runtime/client/watcher.go

Line 183 in f3b5c49

// Check if any transactions are due for a retry.

, since when a message with same payload is re-published the libp2p client doesn't actually publish the message if has already published a msg with that id (this check is done locally before publishing messages, not on the receiving end).

Note: client also retries on epoch transitions, which results in new message payloads (groupVersion changes) which works fine.

So we can either:

• remove the periodic client retry (retry on epoch transitions is needed and should remain)
• update the Tx message format (or hash function) to avoid duplicate messages in this case

In any case disabling retries in E2E tests was actually a good thing as it discovers issues that can otherwise get masked by retrying.