sim report: Avg win (R) / Avg loss (R) print a measured-looking 0.000 for an empty side #1004

Found while fixing #1000 (PR #1003).

keel/sim/report.py edge table renders avg_win_r / avg_loss_r through _r_cell. stats._r_aggregates writes 0 for an empty side of the R sample (field comment: "0 when the R sample has none of that side, the money fields' convention"), so an all-losses row prints Avg win (R) = 0.000 and an all-wins row prints Avg loss (R) = 0.000, both of which read as measured values.

Since #965 the track record uses None for these (_side_average, decided by counting). The edge table could do the same thing, with n_wins_r / n_losses_r deciding whether a side is empty. It could print a bare n/a or reuse _empty_side's no wins yet / no losses yet through undefined_reading. The engine values stay unchanged (pool_stats multiplies by them).

#1000 was limited to the profit factor, as its issue body asked.