(#393) DefaultNumberSystem: slight performance improvements

[andi-huber]

Improvements:

• [x] Avoid creation of stacktraces in boolean isInteger(NumberType numberType, Number number)
• [x] Avoid creation of stacktraces in Number narrow(Number number)
• [x] Add a testcase for BigDecimal like 1234.000 (note the trailing zeros)
• [x] Refactor isZero(Number) check into NumberType, and provide optimized variants for RationalNumber, BigInteger and BigDecimal
• [x] Shortcut multiplication in multiplyWideAndNarrow(...) if any of the operands is zero.

---

This change is 

[keilw]

I was away for a few days, but will have a look tomorrow.

[andi-huber]

Found a view more optimization candidates today. Implemented these. Now PR is ready for review.

[andi-huber]

Why are some methods like

        @Override boolean isZero(Number number) {
                return ((AtomicInteger) number).intValue() == 0;
            }
        },

casting directly to e.g. AtomicInteger while other methods check for instanceof before?

Hi Werner. If we do the isZero(Number number) checks within the (private) enum NumberType, an instanceof check is no longer required, because we already know what type the number parameter has. This is implicit knowledge, which I did not have a good idea how to communicate better within the code or comments.