* 16 bits fix encoding

* Types modification

* Rounding loss check

* Clean up a bit and add benchmarks

Unneded usage of big.Int has been removed

Conversion methods are now methods/constructors of Float16 to make the code
more idiomatic.

Co-authored-by: Toni Ramírez <toni@iden3.com>
Co-authored-by: arnaucube <root@arnaucube.com>
Co-authored-by: Eduard S <eduard@iden3.io>