[FEATURE] use uint8 (16?) to float conversion trick from Elliot Gorokhovsky

[DenisYaroshevskiy]

@embg suggested to do the following

uint8_t-> float

2^23 | my_num == 2^23 + my_num
Then

conversion is (2^23 | my_num) - (2^23)

He is not the first person to do it but who was the first is unknown at the moment.

So far the assumption is - he will put up a diff.