Improve performance of inplace additions

[tonybaloney]

This PR merges the two loops inside v_iadd. For exponents, this gives about a 10% speed improvement.

./python.exe -m timeit -n 100 -r 10 "2**1234567"
100 loops, best of 10: 5.01 msec per loop
# after
./python.exe -m timeit -n 100 -r 10 "2**1234567"
100 loops, best of 10: 4.53 msec per loop