ENH: use math.factorial for exact factorials

[ewmoore]

math.factorial is substantially faster than the naive multiplication algorithm.

In [89]: timeit special.factorial(100, True)
100000 loops, best of 3: 9.1 µs per loop

In [90]: timeit math.factorial(100)
1000000 loops, best of 3: 1.43 µs per loop

Ideally i think that special.factorial wouldn't be a thing, but since it is, there isn't any good reason to use a poor algorithm when a better one already exists.

When n isn't an integer, this will still raise a ValueError, but the text will be different.

[argriffing]

Looks good to me if it passes the CI. According to the python docs math.factorial() was new in 2.6 so this should be OK.

[codecov-io]

@@            master   #5693   diff @@
======================================
  Files          234     234       
  Stmts        43096   43092     -4
  Branches      8154    8152     -2
  Methods          0       0       
======================================
- Hit          33410   33408     -2
+ Partial       2605    2604     -1
+ Missed        7081    7080     -1

Review entire Coverage Diff as of f6012da

Powered by Codecov. Updated on successful CI builds.

[rgommers]

LGTM too, merged.

[endolith]

The docstring says If n<0, the return value is 0. but this is no longer true:

factorial(-1, True)
Traceback (most recent call last):
...
ValueError: factorial() not defined for negative values