Boxoban incremental

[TBBristol]

Adds a new incremental difficulty (5) which is based on maze_incremental
Standard difficulties remain the same and tested
Targets hit is now a % on all difficulties
When a map is solve it increases difficulty up to level 3 (hard) and does not end episode
Keeps loading hard maps if hard is solved
Episode tick is now total ticks, puzzle tick is tick final puzzle
Timeout is 150 per puzzle not episode and based on puzzle tick
For incremental mode scoring is as follows:
score = 0..4 for highest difficulty cleared in the episode (note +1 offset from difficulty labels)
perf = score / 4.0
So if all envs are reaching hard perf is 1

[TBBristol]

perf changed to

perf = (score + targets_hit) / 5.0f;

Where targets his is a % of targets on the level.
This normalises to at all targets hit on HARD but allows for incremental gains where the next level is not yet achieved.

Score is still level achieved.