fix(ReforceXY): reduce PBRS defaults to prevent reward exploitation

fix(ReforceXY): reduce PBRS defaults to prevent reward exploitation

Disable hold potential by default and reduce additive ratios to prevent
the agent from exploiting shaping rewards with many short losing trades.

Changes:
- hold_potential_enabled: true -> false (disabled by default)
- hold_potential_ratio: 0.03125 -> 0.001 (reduced when enabled)
- entry_additive_ratio: 0.125 -> 0.0625 (halved)
- exit_additive_ratio: 0.125 -> 0.0625 (halved)

These conservative defaults encourage the agent to focus on actual PnL
rather than gaming intermediate shaping rewards.