Add compatibility tests with nasm

[postmodern]

Add specs that test whether Program#assemble can output the same bytes as nasm would given the same ASM syntax. This should catch any remaining edge-cases in X86::Encoder and X86_64::Encoder with respect to exotic features, such as:

• SIB memory addressing
• RIP-relative offsets (ex: _label[rip])
• relative jumps
• jumping to a register
• EVEX encoding
  • vector register operands
  • {k}/{k}{z} write masks
  • scalar-to-vector memory broadcasting
  • {er}/{sae}

[postmodern]

It appears that nasm does not support certain instruction names that ronin-asm supports, which are from Opcodes:

BITS 32
section .text
_start:
	aor	[ecx],	ecx

/tmp/ronin-asm-x86-20250318-77905-vmet9i.s:4: error: parser: instruction expected

nasm also appears to have odd implicit operand casting rules that require you omit the BYTE/WORD/DWORD/QWORD size specifier for single operand instructions:

BITS 32
section .text
_start:
	clflush	BYTE [ecx]

/tmp/ronin-asm-x86-20250318-77905-w14n51.s:4: error: mismatch in operand sizes

Also noticing slight differences in byte encoding of instructions:

  2) Ronin::ASM::X86::Instructions::ADD#encode when given operands of types mem32, imm8 is expected to eq "\x80\x01A"
     Failure/Error: expect(output.string).to eq(nasm(subject))
     
       expected: "\x80\x01A"
            got: "\x83\x01A"
     
       (compared using ==)
     # ./spec/x86/instructions/add_spec.rb:415:in `block (4 levels) in <top (required)>'

[postmodern]

Going to skip this due to nasm's encoded instructions being slightly different from our encoded instructions.