math.c: fix tgamma on mingw

	* math.c (mingw_tgamma): tgamma(3) on mingw returns a NaN for a
  big number, not infinity.
	* math.c (ruby_tgamma): fix tgamma(-0.0) on mingw.
	  [ruby-core:74817] [Bug #12249]

	* math.c (ruby_lgamma_r): fix lgamma(-0.0) on mingw and OSX.

	* math.c (ruby_lgamma_r): mswin's lgamma_r also seems to be wrong.
	  cf. [Bug #12249]

	* math.c (ruby_lgamma_r): missing/lgamma_r.c is used on Windows,
	  since msvcrt does not provide it.

	* missing/lgamma_r.c (lgamma_r): fix lgamma(-0.0).
	  [ruby-core:74823] [Bug #12249]

	* configure.in (rb_cv_lgamma_r_m0): check if lgamma_r(-0.0)
	  returns negative infinity.  [Bug #12249]

	* math.c (ruby_lgamma_r): define by the configured result.

	* configure.in (rb_cv_lgamma_r_m0): fix the condition for
	  lgamma_r(-0.0).  [Bug #12249]


git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/branches/ruby_2_2@54687 b2dd03c8-39d4-4d8f-98ff-823fe69b080e