gdb: potential bug due to double address translation

In `add_sw_breakpoint` and `remove_sw_breakpoint`, the address is translated from GVA to GPA using translate_gva, but then the resulting GPA is passed to read_addrs/write_addrs which internally call translate_gva again, treating the already-translated GPA as a GVA. 

I would expect this to cause incorrect memory access since the physical address gets translated a second time?