glib-2.0: make float and double MIN consistent with integers and add MIN_POSITIVE
ANSI C's FLT_MIN and DBL_MIN are the 'minimum normalized positive value'. These are currently bound as float.MIN and double.MIN, but this is inconsistent with the binding for MIN of integers. The inconsistency was highlighted in #680
This patch proposes making float.MIN and double.MIN consistent with Vala's integer bindings and adding float.MIN_POSITIVE and double.MIN_POSITIVE as the new bindings to FLT_MIN and DBL_MIN.