Fix keepalive handling (various issues in original issue tracker)

[uschindler]

Hi,
the original client has some problem with the state machine, which handles the keep alive packets. If the connection breaks and the client is waiting for a ping response, a reconnect with the same client will still wait for the pong response, as the state machine says pingOutstanding is true.

This code cleans it up a bit:

• on connection it makes sure that pingOutstanding=false (that's the main issue)
• it only sets it to true if it successfully sent the ping request
• when the connection is closed, the pingOutstanding state variable is also reset to false for safety
• it also makes sure that the timer is reset when publishing new topics and other places.

This fixes the issues mentioned in the original issue tracker:

• KeepaliveTimeout error with different brokers knolleary/pubsubclient#795
• No Mqtt pings on ESP32 with 2.8. knolleary/pubsubclient#829
• [MQTTPINGREQ] lastOutActivity is not set when replying to ping req. knolleary/pubsubclient#1059
• mqtt client exceeds timeout and disconnects knolleary/pubsubclient#810

This is mainly an improvement to the following pull request on original repo: knolleary#802

It would be nice to include this together with the urgent fix to unbreak passing a password on connect.

I am testing this at moment together with the PR #11 with the device BSB-LAN (for heaters): fredlcore/BSB-LAN#683

[uschindler]

@imbeacon: it would be good to merge this and especially #11.
@MathewHDYT: maybe have a look, too

[MathewHDYT]

@imbeacon Looks good this should definetly be merged with #11 as soon as possible.

Thanks a lot for the improvements @uschindler.