Hi,

   Here is another attempt to fix this issue by emitting
DW_AT_bit_size and DW_AT_bit_offset for float types which have
precisions smaller than their storage size.  Please review.

-Doug