Improve Timeout Information For Server Time Budgeting

[tomdeering-wf]

Today, a server can check the Timeout() on the FContext it receives in order to estimate how much time is left to serve the request. However, this can be a significant overestimate due to:

1. Network latency
2. Time spent sitting in the server's request buffer

It would be great if Frugal could add (1) a best-effort wall time when the client request was made, (2) the amount of time a message spent in the queue, or both. This would help Frugal servers make better estimates of the true client time budget.

[stevenosborne-wf]

https://jira.atl.workiva.net/browse/MSG-448

[brianshannan-wf]

It seems like the two options here are:

1. Have clients put a deadline in the FContext before they make a request (either in addition to or in replace of a relative timeout). One flaw here is clock drift from browser clients could be excessive
2. Have the server deserialize the FContext, put a deadline in from the relative timeout, then reserialize the FContext (ignoring network latency, but would catch time in the queue).

[joshhayes-sheen-wf]

This is something we just ran into too