`dataCount` is decremented above and equals zero at this stage.
The correct way is to use `bufCount` instead.

Co-authored-by: Colin Ian King <colin.king@canonical.com>

Refactor following test modules to use new API handling
number of passed and failed test cases:

- Clear memory
- Custom operation
- HEC
- Poly 1305
- PON
- SHA
- XCBC
- SNOW3G
- ZUC

Fixed issue due to multiple definitions of "imb_errno".

- basic API for test suite management added
  - start and end test suite, counting pass/fail test cases
  - structured and consistent result print format

When passing multiple buffers with different buffer sizes,
the common length of them would be processed, as in CBC mode.
The problem is that the input/output pointers will be moved
equally in all buffers, meaning that the pointer of
a non-completed buffer might be pointing at a block that should
not be encrypted, and therefore, it will be encrypted in the next call.

Instead of this, the pointers are moved in steps of 160 bytes,
making sure that for non-new buffers, the input/output pointer
points at the next block to be encrypted

When generating keystream to authenticate multiple buffers,
only 8 bytes more of keystream are needed when all data
to authenticate has been processed, so only 8 more bytes
were generated for all buffers.
The problem here is that this only works if all buffers
have the same size. If the other buffers have different
sizes, they will need 16 bytes to be generated before
getting into the single buffer code path.

Fixed the byte mask when index is 0, which also is used
to read/write full 64 bytes of plaintext.

Also fixed clamping to have length multiple of 4 bytes,
but for lengths higher than 255 bytes too.

CBCS requires a message length multiple of 16 bytes,
so a check needs to be added.

Handle up to 4 Chacha20 states, reducing the number of
shuffles needed.

Handle up to 2 states at a time, to maximize CPU utilization.