ITS 1254, core dump in numericStringNormalize()

[Peter W <peterw@usa.net>]

Quick notes I was unable to add to the ITS ticket:
- suggested fix does not work because...
- isn't (q <= p) an assertion about logical memory addresses?
- does it mean to assert something like newval->bv_val being no shorter than
  val->bv_val
- which would be "assert( (q - newval->bv_val) <= (p - val->bv_val) );", 
  right? That assertion *instead* of the two current assertions, right?
- shouldn't the code ensure that the ch_malloc() calls succeeded?

Thanks,

-Peter