v2.5.1: print-start gate uses filament_exist, not the idle motion sensor

[bbolinger]

Print-start gate no longer refuses a loaded-but-idle tool

The preflight check that confirms filament is loaded before a print starts was reading the tool's motion sensor (filament_motion_sensor.filament_detected). That sensor only reads true while filament is actively moving through a feeding head. At the moment you approve a print every tool is idle, so a genuinely loaded tool (filament_exist=True, visible colour/material) read as "not loaded" and the start was refused.

Verified against the live U1: a real PLA print on T3 refused with exists=True, motion_detected=False, feed_detected=True. With the fix the same case passes; a genuinely empty tool (filament_exist=False) still blocks.

Change

• u1_toolmap.summarize gates on filament_exist (the printer's per-tool loaded/not-loaded flag) instead of the motion/feed sensors. The motion sensor is the printer's own runout mechanism during a print, not a pre-start loaded check.
• Regression tests: idle loaded tool (exists=True, motion/feed False) must not block; empty tool (exists=False) must block.

Latent until now because prior drills always ended at grace-cancel, never reaching a real (non-cancelled) start where this preflight fires.

Verification

Full suite: 1018 passed, 8 skipped. Reproduced the exact live failing case (PLA/T3) and confirmed the gate now allows it while still blocking an empty tool.

[bbolinger]

Closing: this change gated the start on filament_exist, which is a software-assigned flag (set via the UI / POST /printer/filament_detect/set), not a physical sensor. It read True both while the filament was physically loaded AND while it was parked/unloaded, so gating on it would let a not-actually-loaded tool through (fail-open). The original gate reads the per-channel filament_motion_sensor (a hardware switch at each feed, pin eN:PA1), which correctly reflects physical presence. The original fail-closed gate is correct; reverted on the box.