Full big num implementation with preferred serialization and 65-bit negs

[laurencelundblade]

Encoding and decoding of negative big numbers now takes into account the offset of 1 for all CBOR negative numbers. Also, big numbers are encoded with preferred serialization -- when they can be encoded with type 0 and type 1 integers, they are. Encoding and decoding big numbers is no longer a pass through for tagging a binary string.

This is an incompatible change with QCBOR 1.x. A mode configuration call is added to return to the 1.x behavior if desired.

This is because of the realization in work on 65-bit negative numbers that big numbers need preferred serialization.

This affects big floats and decimal fractions when the mantissa is a big number.

New methods to encode big numbers with non-preferred serialization are added.

A new method is added to process a raw big number for the offset of one for negatives. This is outside of spiffy decode. It does a little big number arithmetic.

dCBOR numeric reduction now includes 65-bit integers both for number encoding and decoding.