Improve _check_key() and _store_cmd() performance

[shargan]

While making the batching changes in #182 I noticed a few performance hot-spots.

1. _check_key() is called at least once on every single operation and was doing a lot of repeat or just inefficient work.
2. _store_cmd() was calculating and encoding extra and expire for every key/value pair even though they don't change (I didn't move this in the previous PR because I didn't want to muddy my intentions).

The first change is responsible for most of the performance increase, it affects every method, and the difference scales with key length. The second only impacts set_many() but in my 100-key tests was responsible for a decent 7.5% gain.

Benchmarks from master with irrelevant lines trimmed:

# python2.7
$ pytest --verbose --capture=no --no-cov -m benchmark pymemcache/test --keys 100 --count 10000
================================================ test session starts =================================================
platform darwin -- Python 2.7.15, pytest-3.7.4, py-1.6.0, pluggy-0.7.1 -- /Users/shargan/repos/pymemcache/.venv/bin/python

pymemcache/test/test_benchmark.py::test_bench_get[pymemcache] 2.37730312347
pymemcache/test/test_benchmark.py::test_bench_set[pymemcache] 0.353810071945
pymemcache/test/test_benchmark.py::test_bench_get_multi[pymemcache] 16.3643598557
pymemcache/test/test_benchmark.py::test_bench_set_multi[pymemcache] 10.8592970371


# python3.6
$ pytest --verbose --capture=no --no-cov -m benchmark pymemcache/test --keys 100 --count 10000
================================================ test session starts =================================================
platform darwin -- Python 3.6.6, pytest-3.7.4, py-1.6.0, pluggy-0.7.1 -- /Users/shargan/repos/pymemcache/.venv36/bin/python3.6

pymemcache/test/test_benchmark.py::test_bench_get[pymemcache] 3.1345767974853516
pymemcache/test/test_benchmark.py::test_bench_set[pymemcache] 0.3885829448699951
pymemcache/test/test_benchmark.py::test_bench_get_multi[pymemcache] 18.19443702697754
pymemcache/test/test_benchmark.py::test_bench_set_multi[pymemcache] 11.366436004638672

With this diff:

# python2.7
$ pytest --verbose --capture=no --no-cov -m benchmark pymemcache/test --keys 100 --count 10000
================================================ test session starts =================================================
platform darwin -- Python 2.7.15, pytest-3.7.4, py-1.6.0, pluggy-0.7.1 -- /Users/shargan/repos/pymemcache/.venv/bin/python

pymemcache/test/test_benchmark.py::test_bench_get[pymemcache] 2.65929102898
pymemcache/test/test_benchmark.py::test_bench_set[pymemcache] 0.332684993744
pymemcache/test/test_benchmark.py::test_bench_get_multi[pymemcache] 12.0730199814
pymemcache/test/test_benchmark.py::test_bench_set_multi[pymemcache] 5.42873597145


# python3.6
$ pytest --verbose --capture=no --no-cov -m benchmark pymemcache/test --keys 100 --count 10000
================================================ test session starts =================================================
platform darwin -- Python 3.6.6, pytest-3.7.4, py-1.6.0, pluggy-0.7.1 -- /Users/shargan/repos/pymemcache/.venv36/bin/python3.6

pymemcache/test/test_benchmark.py::test_bench_get[pymemcache] 2.7322170734405518
pymemcache/test/test_benchmark.py::test_bench_set[pymemcache] 0.32326221466064453
pymemcache/test/test_benchmark.py::test_bench_get_multi[pymemcache] 11.463497877120972
pymemcache/test/test_benchmark.py::test_bench_set_multi[pymemcache] 6.002143859863281

[cgordon]

Nice improvement, thank you!