PERF: various performance optimizations

[jreback]

PERF: enhance tseries getitem indexing
PERF: micro optimzations in inference
API/CLN: core.internals.apply now only takes kwargs; makes code a bit simpler
CLN: wrap core/common/isnull into more direct Block.isnull
PERF: fast path series block operations

closes #6264

original timings

In [11]: %timeit com.is_integer(13)
100000 loops, best of 3: 1.29 µs per loop

In [12]: %timeit com.is_integer(np.int32(2))
100000 loops, best of 3: 3.32 µs per loop

In [13]: %timeit com.is_integer(np.int64(2))
100000 loops, best of 3: 1.86 µs per loop

In [14]: %timeit com.is_integer(14L)
100000 loops, best of 3: 1.29 µs per loop

In [15]: %timeit com.is_integer("1")
100000 loops, best of 3: 2.2 µs per loop

withi this PR

In [2]: %timeit com.is_integer(13)
10000000 loops, best of 3: 72.5 ns per loop

In [3]: %timeit com.is_integer(np.int32(2))
1000000 loops, best of 3: 620 ns per loop

In [4]: %timeit com.is_integer(np.int64(2))
1000000 loops, best of 3: 315 ns per loop

In [5]: %timeit com.is_integer(14L)
10000000 loops, best of 3: 72.4 ns per loop

In [6]: %timeit com.is_integer("1")
10000000 loops, best of 3: 78.8 ns per loop

[cancan101]

typo here?

[jreback]

yep....suprising didn't break anything (changed though!) thanks

[jreback]

closing in favor of #6269