V0.3.7 patch21 extra tests around demurrage

[benjaminbollen]

still significant bug in ERC20 Inflationary

1. there was additional logic to absorb conversion errors between demurrage and inflationary format, but that implementation itself had a bug. It SHOULD have been of the form

- Extend numerical precision
- do the computation
- trash the extra Least Significant Bits we added

but instead it was wrongly implemented as

- Extend the numerical precision
- do the computation to convert to inflationary format
- don't yet trash the extra least significant bits

then on unwrapping

- do the computation to convert back
- trash the extra bits

this actually caused the extra bits to pollute the other bits more, introducing a bigger inaccuracy than without this extra complexity.

In hindsight the obvious insight is that GAMMA and BETA have a fixed precision, so multiplying by them keeps the same amount of accurate bits, regardless of shifting the bit registry, so the whole exercise just complicated the code and costs extra gas.

So it is removed.

2. also removed some extra dead / redundant code in demurrage

[benjaminbollen]

duplicates of public convertDemurrageToInflationaryValue

[cducrest]

Looks fine