Wrong evaluation of i386 instructions

Background: http://kronotai.com/wordpress/339/testing-decompiler-semantics/
Subject: https://github.com/rfalke/decompiler-subjects/tree/master/from_holdec/stress_int_ia32/ia32_elf_base_inst__1_var__no_complex

The test functions for the following instructions do not return the expected zero:
- `cmpxchg`: inst_94_values_var_0()
- `rol`: inst_159_values_var_0(), inst_164_values_var_0(), inst_169_values_var_0(), inst_174_values_var_0()
- `ror`: inst_165_values_var_0(), inst_175_values_var_0()
- `sar`: inst_166_values_var_0()
- `rcr`: inst_210_values_var_0()

Also the carry flag is incorrectly calculated for `sbb`: inst_221_flags_var_0(), inst_224_flags_var_0(),inst_227_flags_var_0(),inst_228_flags_var_0(),inst_229_flags_var_0()

Start with the linked decompiler-subjects test case and the named functions: inst_94_values_var_0(), inst_159_values_var_0() through inst_175_values_var_0(), inst_210_values_var_0(), and the listed sbb flag tests. Run those tests to reproduce the incorrect instruction results and carry flags. Done means the affected functions return zero and sbb carry flags match the expected values.

Written by the indexing model from the issue text.