snode communication and logging clean up #2 & AppDotNot minor refactor

[neuroscr]

overall:

• make sure NODE_TLS_REJECT_UNAUTHORIZED is set correctly with the research Jeff did

snode stuff:

• increase random node pool size to 1024 to match mobile (trello has a card on this)
• sendToProxy now uses agent instead of the TLS global hack
• handle 401s with a retry, handle 500s with a delayed retry
• make sure getRandomSnodeAddress can't stack
• minor refactoring and internal function renaming
• logging improvements
• allow http-resource pollServer disconnect to be correct even if it's been stopped before

adn stuff:

• refactor functions to be useable outside of class, make it easier to integrate with proxy/onion library in snode stuff
• .loki support improvements
• remove pollForChannelOnce loop, rely on timer instead

[majestrate]

ideally the code should never ever ever touch NODE_TLS_REJECT_UNAUTHORIZED and should instead properly handle custom tls certs, or alternatively permit http for .loki gtld.

[msgmaxim]

The name doesn't seem to suggest that it would return something. I see that randomSnodePool.lenght is only used for logging, why not log inside markRandomNodeUnreachable? Like "Marked {} as unreachable, {} left in the pool". Then you can have another log message one level above telling the reason (which can be different from what I can see).

[neuroscr]

it mainly to reduce the number of logging messages. I had that way initially but we already log everytime this is called as it's a p. big deal event and the outer context is more important that the fact a node was removed but it is VERY helpful to know how main are left to ensure roll overs are happening as expected.

[msgmaxim]

If you don't want two separate log messages, how about we make another call to getRandomSnodePool() to get the size? (Or add another method to just get the size, without potentially calling initialiseRandomPool.)

[neuroscr]

well both actions need to happen the banning of the node and getting the remaining count. Is it the name of the function that is the issue?

A compromise maybe the two function solution but seems a bit inefficient, since banning the node needs to happen and knows the count. But I guess it is just a member variable it needs to count, so not as bad as the unreachableNode node situation.

I'll just split for now. Due to the time crunch of trying to get this in for 1.0.5, I'll leave a temp variable in for now and clean it up later.

[msgmaxim]

Don't we need to reset this.initialiseRandomPoolPromise here?

[neuroscr]

Yes

[Mikunj]

minor things

[Mikunj]

Looks good