Fixing arbitrary 40 digit max precision in printf code

Prompted by http://www.exploringbinary.com/print-precision-of-dyadic-fractions-varies-by-language/ I agree with the author that 40 is a weird arbitrary limit to impose here. As he writes, the smallest float that will fit in a double is 1074 decimal digits and the largest is 53. I don't see a reason we shouldn't be in line with Perl and C here. It's a couple of minor constant size changes in main/snprintf and ext/standard/formatted_print.c plus a test case fix. -Rasmus