Fix packed x86

[gitoleg]

This PR brings a lot of changes in x86 packed instructions. The main goal is to make a bil code closer to the BIL normal form, i.e. to reduce an amount of an additional code injected during the normalization process as well a time spent for lifting.
As a result, bap works better with files that previously took approximately infinite time.

Affected instructions are:

• different cases of pcmp, pmin, pmax;
• instructions like psrl, psll, pavg, xorp;
• instructions that compare strings: pcmpistri

Just few numbers to compare:
it was:

$: echo "\x66\x0f\xda\xc8" | bap-mc --show-bil | wc -l
327678
$: echo "\x66\x41\x0f\x3a\x63\x08\x4a" | bap-mc --show-bil | wc -l
Fatal error: out of memory.

it is:

$:echo "\x66\x0f\xda\xc8" | bap-mc --show-bil | wc -l
132
$: echo "\x66\x41\x0f\x3a\x63\x08\x4a" | bap-mc --show-bil | wc -l
139

Also, This PR adds few instructions to x86-lifter:
pmaxsb, pmaxsw, pmaxuw, pmaxud, pmaxsd and tests for them.

[ivg]

👍 great job, thanks!

[ivg]

try to use open Core_kernel in the new code, the Std module is deprecated