fix(IRSB): removed wrong instruction_addresses calculaion from lift…

[DanielBotnik]

I had one of the weirds bug while using pyvex on one of my binaries, I found out that the irsb don't count addresses after empty IMark for some reason, but letting it recalcualtes it works fine.

In [1]: import pyvex

# MOV R0, 0
# SEV
# SEV
# MOV R1, 3
In [2]: opcodes = b"\x03\x00\xa0\xe3\x04\xf0\x20\xe3\x04\xf0\x20\xe3\x03\x10\xa0\xe3\xfa\x0b\x00\xea"

In [3]: irsb = pyvex.IRSB(data=opcodes, mem_addr=0x2000, arch=pyvex.ARCH_ARM_LE)

In [4]: [hex(i) for i in irsb.instruction_addresses]
Out[4]: ['0x2000', '0x2004']

In [5]: irsb.pp()
IRSB {
   t0:Ity_I32 t1:Ity_I32 t2:Ity_I32

   00 | ------ IMark(0x2000, 4, 0) ------
   01 | PUT(r0) = 0x00000003
   02 | ------ IMark(0x2004, 0, 0) ------
   03 | ------ IMark(0x2004, 4, 0) ------
   04 | ------ IMark(0x2008, 4, 0) ------
   05 | ------ IMark(0x200c, 4, 0) ------
   06 | PUT(r1) = 0x00000003
   07 | ------ IMark(0x2010, 4, 0) ------
   NEXT: PUT(r15t) = 0x00005000; Ijk_Boring
}

In [6]: irsb._instruction_addresses = None

In [7]: [hex(i) for i in irsb.instruction_addresses]
Out[7]: ['0x2000', '0x2004', '0x2004', '0x2008', '0x200c', '0x2010']

Now I also noticed that there was multiple addresses, so I made sure only to include IMARK with len greater than 0. this seemed to solve the problem