Support return commands from previous decision to save overhead of calling waitUntil

[longquanzheng]

A lot of the waitUntil api is quite simple —- just return a static command request. For those cases, it’s more efficient to return them from the previous state decision

[longquanzheng]

One idea is to let the state decision to call the next state ‘a waitUntil api in advance (it even has the input already).

And then provide a flag in workflow state option to opt in this optimization.

We may provide an IDE plugin to hint user to use this feature— based on the implementation of waitUntil

[longquanzheng]

WaitUntilEagerly , not persistence or communication. User code shouldn’t call external api

[longquanzheng]

The implementation will be similar to

iwf/service/interpreter/workflowImpl.go

Line 546 in 34052f7

if len(commandReq.GetInterStateChannelCommands()) > 0 {