Revert default behavior change with data publishing

[davidzhao]

No description provided.

[dennwc]

I think the expectation about using server SDK is that it abstracts away the packet drop and just delivers app-specific data reliably by default. I think setting it as unreliable is a kind of optimization that app developers may consider (later?). This is why I changed it this way.

Also note that in our own protobuf the default is also RELIABLE. So it also matches the generated code better. One less quirk when generating new SDKs.

So I'm wondering why you'd like the default to be UNRELIABLE. Is it expected by app developers?

[dennwc]

From Slack:

There are two reasons for publishing as lossy by default:

1. consistency - all of our real-time SDKs publish unreliable delivery by default. it would be unexpected for Go to behave differently
2. semantics of reliable delivery - we've been trying to move away from "reliable" delivery because it's not actually reliable especially in the mesh setting. since we do not offer delivery guarantees, it's confusing for users to expect "reliable" delivery when they are getting "best-effort" delivery.