Consider tweaking httpMaxRetryDuration and httpRequestTimeout

[SimonWoolf]

Currently httpMaxRetryDuration is 10s, which is less than httpRequestTimeout. That means that if a library takes TO3l6 at its word, it contradicts RSC15d: the latter requires a fallback host attempt on http timeout, but that's barred if httpMaxRetryDuration < httpRequestTimeout.

(ably-js currently avoids this by not taking httpMaxRetryDuration seriously, but we should decide on a proper solution for 1.0/1.1/whatever. See previous discussion re ably-js here)

[mattheworiordan]

Given the insignificance of this change, I think we should fix it now.

Perhaps just as simple as 12s for an HTTP request and max retry duration of 15s?

[SimonWoolf]

Sounds ok to me (or maybe 10s http / 15s mrd, or 12s http / 18s mrd, to give the second request a little more breathing room). @paddybyers might have an opinion on this re good interactions between clientside timeouts and realtime timeouts (see prev discussion on those here)?

[paddybyers]

I tend to agree that 12/15 doesn't give enough headroom for subsequent requests.

I'm not sure there's really much difference between 10/15 or 12/18; probably 10/15 is best. I don't think we get many requests succeeding in the 10-12s window.