Calculate exp, log, pow with the given prec even if prec limit is set

[tompng]

BigDecimal.limit(1)
BigMath.exp(123, 100) # This should return 100 digit precision value.
# Of course internal calculation uses precision larger than 100.

Wrapped with save_limit

BigDecimal.save_limit do
  BigDecimal.limit(0)
  ...
end

Compact diff: https://github.com/ruby/bigdecimal/pull/411/files?w=1

sqrt has the same problem. It will be fixed in #381