PnL calculation had been change from 'Per Trade' evaluation to 'Per Filled Order' (AKA sub trade).
The motivation was to make the reward denser, by introducing the realized pnl as soon as possible (next skip_frame)
without having to wait for the trade to end (which is indicated by position crossing zero and is undetermined in time)
Unrealized PnL calculation had been redesign. now every position being held is evaluated by price change of previous bar.
Broker stat and variables that wasn't used/became redundant had been removed
Broker Stat Rework: (https://github.com/Kismuz/btgym/issues/117)