Pedantic: Extend byte_length()/bit_length() to support systems where CHAR_BIT is not 8

[saxbophone]

Current code assumes 8-bit (a reasonable assumption):

arby/arby/include/arby/Nat.hpp

Lines 827 to 840 in 8f9b0bd

	std::size_t leading_occupancy = PRIVATE::fit(_digits.front(), 256);
	bytes_for_digits -= (sizeof(StorageType) - leading_occupancy);
	return bytes_for_digits;
	}
	/**
	* @returns size by number of bits needed to store the number's value
	* @note this can be less than \f$ bytes \times 8 \f$
	*/
	constexpr std::size_t bit_length() const {
	// this is how many bits are needed to store the digits
	std::size_t bits_for_digits = _digits.size() * sizeof(StorageType) * 8;
	// reduce size if leading digit is not full occupancy
	std::size_t leading_occupancy = PRIVATE::fit(_digits.front(), 2);
	bits_for_digits -= (sizeof(StorageType) * 8 - leading_occupancy);

But we can make this portable without compromising the behaviour on the vast majority of systems where CHAR_BIT=8 by using CHAR_BIT in place of 8 and PRIVATE::exp(2, CHAR_BIT) in place of 256

This should support any systems where CHAR_BIT > 8. The rest of the code should be checked over for any other places where byte = 8-bit is assumed.