Retry blocking client write

[dgreatwood]

Added simple extra method to retry synchronously client-side writes that would otherwise have blocked. Note: A more normal technique would be to do a retry on write coming available in the socket (fd) poll loop, but queueing client writes would require more changes to the core Pistache code, so for now can simply to do the retry directly.

Note: Without this change, protocol interchanges will hang as soon as a would-block error occurs; in my testing within a minute or two of beginning to communicate actively on the socket.

[dennisjenkins75]

Hello. This PR has some merge conflicts. The pistache client has not received much love, so PRs to improve it would be very welcome. If you are still interested in having this one merged, please refresh it. Otherwise, I'll likely close this PR in a few weeks (we're doing a PR cleanup).