…se code than

Job=ShortRun  IterationCount=3  LaunchCount=1
WarmupCount=3

| Method                     | a | b          | Mean     | Error     | StdDev    | Allocated |
|--------------------------- |-- |----------- |---------:|----------:|----------:|----------:|
| Mul64By32_New | 3 | 4294967295 | 2.068 ns | 0.0459 ns | 0.0383 ns |         - |
| Mul64By32_Ori                |  3   | 4294967295 | 2.916 ns | 0.0231 ns | 0.0193 ns |         - |

- Add comment to BigMul64By32 and make it return nunit to avoid clearing upper 32 bits
- Simplify IncreaseScale

…it can easily be removed once JIT recognize and optimize "ulong * uint"

…ding

* Gives around 20% faster perf for 64bit for full 96bit division
* removes cast

* makes the 5-10% regression 20% faster (25% speedup) for 1/3

* Gives up to 10ns (or 14% faster)

… code

…ental]