py/formatfloat: Add ability to format doubles with exponents > 99.
For single prec, exponents never get larger than about 37. For double prec, exponents can be larger than 99 and need 3 bytes to format. This patch makes the number of bytes needed configurable. Addresses issue #1772.