Implement multi-buffer version of the ZUC-EIA3 algorithm,
making use of the 4-buffer keystream generation and
initialization functions.

Add tests to verify that different buffers, with different keys,
lengths, etc. are encrypted correctly using the multi-buffer
ZUC-EEA3 API.

Only print information about individual tests
if DEBUG flag is set, helping reducing the test output.

Unit tests were only checking the first buffer returned
when encrypting/decrypting with ZUC-EEA3.

The code has been refactored, creating a single function
to compare an output buffer with a reference test vector,
used to compare all buffers returned by the ZUC API.

Allocate, initialize and use a single array of buffers
and iterate through it for all the KASUMI F8 N buffer tests,
instead of using an array of array of buffers.

Thanks to this simplification, some issues reported by Valgrind
due to the complex test (including usage of uninitialized variables,
leading to potential segfaults) are resolved.

Add API to perform GHASH on an input buffer,
giving up to 16 bytes of output.

IMB_GHASH() requires gcm_key_data structure to be initialized
calling previously IMB_GCM_PRE_xxx() function, which is the key precompute
function used for AES-GCM, before it is passed as one of the parameters.

- uint128_t renamed to imb_uint128_t

- Convert msg len from bytes to bits

When encrypting/decrypting buffers that are not 64-byte multiple long,
the remaining last bytes (less than 64 bytes) were calculated
by generating 64 bytes of keystream.

Instead, less keystream can be generated (in multiple of 4 bytes,
which is the size of a ZUC word), saving extra cycles.

Implemented keystream generation functions that can
generate from 4 to 64 bytes of keystream, in steps of 4 bytes,
to handle more efficiently less than 64-byte buffers.

Instead of performing single Sbox lookups (for both S0 and S1 S boxes),
in ZUC algorithm, use the new S0 and S1 box computation macros
to compute 16 values in a single XMM register, which is done
in constant time.

Instead of performing single Sbox lookups (for both S0 and S1 S boxes),
in ZUC algorithm, use the new S0 and S1 box computation macros
to compute 16 values in a single XMM register, which is done
in constant time.

Split ZUC initialization and keystream generation
functions, which are scalar only, into SSE and AVX
versions.
This is a preparation for a new commit where
vectorized code will be introduced.

Add macro to compute 16 S1 box values in an XMM register,
which corresponds to 16 S1 box lookups.

Add macro to compute 16 S0 box values in an XMM register,
which corresponds to 16 S0 box lookups.

- small typo fixes