The result of converting the IP address of the broker wasn't checked. As a result, the pointer was left uninitialised and the IPv6 address used for connecting was some random data. The function now returns an error. Before connect_to_broker is called, mqtt_register is executed, which memsets conn to 0, making its state 0 (MQTT_CONN_STATE_ERROR). In order to recover from this error state, the extra check was added in the MQTT_CLIENT_STATE_NEWCONFIG state.
This was discovered using [CodeSonar](https://www.grammatech.com/products/codesonar)
When a connection is aborted by the HTTP server while it's still being processed it is possible to hit a null pointer dereference issue by jumping back to a protothread (outputpt) after its httpd_state has been freed. This can be triggered by sending a POST to any form in the CC26xx web demo server using Firefox.
This patch prevents that by zeroing out httpd_state structs before freeing them, thus also clearing the httpd_state->outputpt field.
Tested using Firefox 55.0a1 on a CC2650 LaunchPad.
The current version of the CC26xx web demo publishes over MQTT the default parent's IPv6 address and the last observed RSSI of this link. This is collected by active probing (periodic ping).
This commit brings the probing functionality to the example's main code module. The MQTT client keeps publishing as previously, but we now also export the same information through CoAP resources. Configuration is still possible through the example's web server.
Instead of using a separate data structure to request that a PD remain powered during deep sleep,
we do the same within the main LPM data structure through an additional field.
This allows us to maintain only one linked list of LPM modules and overall improves code clarity
* Query the sensor about its state, rather than using variables in the driver
* Correctly put the sensor to deep sleep
* Fix doxygen comments
* Don't turn off the sensor in examples since it is no longer needed
