bpo-47127: Specialize calls for fastcall c functions with keywords

[kumaraditya303]

Benchmark:

import pyperf

runner = pyperf.Runner()
runner.timeit(name="benchmark calls",
              stmt="a.decode();a.decode();a.decode();a.decode()",
              setup="a=b'a'*1000")

Result:

Mean +- std dev: [base] 2.16 us +- 0.12 us -> [patch] 1.68 us +- 0.06 us: 1.28x faster

It is the second largest specialization failure as per the stats.

https://bugs.python.org/issue47127

[Fidget-Spinner]

LGTM.