x86: fix decoding of mandatory prefixes

[jxors]

Your checklist for this pull request

• I've documented or updated the documentation of every API function and struct this PR changes.
• I've added tests that prove my fix is effective or that my feature works (if possible)

Detailed description

As mentioned in #2853, the current decoding of mandatory prefixes is not correct. It expects a hard-coded ordering of prefixes that is not required by instruction reference manuals, and also does not handle repeated 66 prefixes.

Initially, I tried to modify the existing function with additional checks. However, it is not possible in general to determine whether a prefix should function as a mandatory prefix without inspecting the opcode, so these checks need to occur later in the decoding process.

To fix this, I have removed the mandatoryPrefix field and replaced it with a function that resolves conflicting mandatory prefixes in the getID function.

Test plan

I have added 23 27 64 tests, comprised of instructions that were originally decoded incorrectly by capstone as well as unexpected edge-cases. For example, instructions with opcode 0FB8..0FBF treat only the REPE/REPNE prefixes as mandatory, and still expect the data size override to function normally.

In general, this patch should only affect instructions in the secondary opcode map with both a REP (F2/F3) and a data size override (66) prefix.

Closing issues

closes #2853

[jxors]

I've marked this as a draft, because 664ef3670f38f62427 is still not handled correctly with this patch.

It used to decode as adcx esp, dword ptr [edi] (which is incorrect). It now fails to decode.

I will fix this issue and do more testing to make sure everything works as intended.