Optimize writing of payload in publish_P()

[hmueller01]

pubsubclient3/src/PubSubClient.cpp

Lines 513 to 515 in a63a531

	for (size_t i = 0; i < plength; i++) {
	rc += _client->write((uint8_t)pgm_read_byte_near(payload + i));
	}

This is rather slow, at least on ESP8266 as this may read every pgm byte 4 times.
Maybe for some other PR to look into making this faster, like how it is done in the esp8266/Arduino String class.
N.B. for ESP32 it will be different, so that's why it is something for a new PR to look into.

Originally posted by @TD-er in #30 (comment)

[TD-er]

For reference, we might implement it like this:
letscontrolit/ESPEasy@66c12eb
Still have to test that code in ESPEasy as it is only used in publishing large messages with beginPublish and endPublish.

[TD-er]

Hmm there has been so many changes in the past 2 weeks, my code can't be used anymore and I simply don't have the time for it to adapt it as I've literally been working for 12 - 15 hours a day the last 3 - 4 weeks....

[hmueller01]

closed by #59