Bug fix: Properly apply gasMultiplier to gasEstimates

[CruzMolina]

Fixes a bad math bug (#2539)!

[coveralls]

Coverage remained the same at 74.12% when pulling e43a5ae on fixGasMultiplier into a158745 on develop.

[gnidan]

QED? (and hope @haltman-at doesn't see my poor formal methods...)

[gnidan]

So, my algebra is rusty, but I had to reason this out in vim for myself to make sense of it, so thought I'd share:

This PR changes this:            To this:

        round( d * 10^P )            n * round( d * 10^P )
v = n * -----------------        v = ---------------------
            10^P                              10^P

We're trying to do this:

v = n * d

Or, expressed as digit placeholders (with underscores as precision gaps):

VVVV.0 = NNNN.0 * D.DDD____

where:

• v is the result we want
• n is bignum,
• d is decimal
• P is significantFigures, i.e. some amount of precision

To [try to?] show how this might play out, the new code does steps:

1. Scale for desired precision (multiply+round):

   D.DDD____  ->  DDDD.0
   

2. Multiply by our bingum:

   NNNN.0 * DDDD.0 = VVVV____.0
   

3. Then divide by original scale:

   VVVV____.0  ->  VVVV.____
   

No loss of precision.

To confirm why the old one was bad, the difference is that we scaled then immediately de-scaled:

1. Scale:

   D.DDD____  ->  DDDD.0
   

2. De-scale:

   DDDD.0  ->  D.0
   

3. Then multiply:

   NNNN.0 * D.0 = VV__.0
   

OK, I'm convinced this is correct and that the old one was not correct.

[gnidan]

(Should this reasoning be a code comment? Would that spare future incredulity?)

[CruzMolina]

I'l leave this convo open for a future Truffler. ⏲