Out of curiosity, do you know how much faster this is?

Tested on the email-enron dataset (after cleanup() with default arguments) for the edges and the size EdgeStat:
_color_arg_to_dict (old): 3.08 s ± 290 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
_color_arg_to_dict (new): 119 ms ± 9.86 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
_scalar_arg_to_dict (old): 2.74 s ± 294 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
_scalar_arg_to_dict (new): 3.19 ms ± 74.5 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

So a 1000x speedup???? Jeez.