reassembler: Fix addition type error in address calculation

[nmeum]

When an instruction contains an operand that is neither a memory nor an immediate operand (e.g. a floating-point operand) then operand_offset is set to None:

angr/angr/analyses/reassembler.py

Lines 726 to 733 in b0d0995

	operand_offsets = []
	for operand in capstone_instr.operands:
	if operand.type == capstone.CS_OP_IMM:
	operand_offsets.append(capstone_instr.imm_offset)
	elif operand.type == capstone.CS_OP_MEM:
	operand_offsets.append(capstone_instr.disp_offset)
	else:
	operand_offsets.append(None)

This causes a type error when calculating the address in register_instruction_reference:

angr/angr/analyses/reassembler.py

Line 1980 in b0d0995

addr = insn_addr + operand_offset

I ran into this issue while packaging angr for Guix where test_ln_gcc_O2 fails because of this:

self = <Reassembler Analysis Result at 0x7fffe4c63190>, insn_addr = 4201104
ref_addr = 6344696, sort = 'absolute', operand_offset = None

    def register_instruction_reference(self, insn_addr, ref_addr, sort, operand_offset):
	if not self.log_relocations:
	    return

>       addr = insn_addr + operand_offset
E       TypeError: unsupported operand type(s) for +: 'int' and 'NoneType'

It seems to me that this regression was introduced in #1739, it is presently unclear to me why this doesn't fail on the CI. Looking at the aforementioned pull request, it seems to me that it was intended to use a zero offset for these instructions. This commit changes the implementation accordingly, thereby fixing the aforementioned test failure for me.

[ltfish]

Thanks for the fix!