Scheme numeric reader of exact exponential constants

As reported by jcowan on LiberaChat's #lisp, the constant number #e1e500 should be read as (expt 10 500), not as (exact 1e500), and should thus produce a bignum. Currently, it errors due to the finite range of double floats. The other exponential letters in §6.2.5 of the R7RS standard such as f for single floats or d for double floats will maintain the old behavior. Thus, #e1.0e500 and #e#1d500 will still overflow (although it is unclear if the first one should). The new behavior must also work when base-10 is specified, i.e. #d#e1e500 and #e#d1e500.