Behaviour commands: reward terms and two probes

Why the first run ignored the gait and swing-height commands was read from its reward terms, and two runs of 200 M steps each changed one reward weight. Neither weight is the fix.

• Terms of the commands that are followed fell by two thirds over the first run: body height −0.188 → −0.065 per step, posture −0.208 → −0.062
• The swing-height term was small from the start: −0.071 per step against +1.231 for velocity tracking
• Probe 1, swing-height weight × 10: swing height 5.4 cm at an 8 cm command (control 4.9 cm), under the 5.5 cm line set before the run
• Probe 2, gait weight × 4: the robot stops when pace, bound or pronk is commanded, 0 of 30 passes on flat ground (control 30 of 30, walked as a trot)