Fix filament accounting with spoolman on mmu setups (happy hare, afc)

### Summary: Rebase E high-water mark on active spool change to prevent spoolman undercount after tool changes ### Issue description: The spoolman integration tracks filament usage using a global “high-water mark” of toolhead E position (_highest_epos). This works for single-material prints, but in MMU (filament swap) scenarios, the E position can be lower than the previous peak at the moment a new spool is activated. This happens due to retract/state-restore/unload sequencing and extruder moves split between macros (for unloading) and direct extruder control in the MMU control software python files. As a result, the first portion of extrusion on the new spool (often ~tens of mm) only brings E back up to the old peak and is therefore not counted, leading to consistent undercounting on the newly activated spool (e.g. ~45mm missing after a 336mm purge). ### Data gathered: By adding the below statement at the def set_active_spool(self, spool_id: Union[int, None]) function we can track whether there is any deviation between the high water mark e position (_highest_epos) and the last observed e position (_last_epos) `logging.info(f"Spool change: new={spool_id}, last_epos={self._last_epos:.3f}, highest={self._highest_epos:.3f}")`  ### Fix: Track the latest observed E position (_last_epos) and rebase _highest_epos to the current E whenever the active spool changes. This ensures usage accounting for a new spool starts immediately from the activation point rather than being blocked by an old E peak. ### Implementation details - Add self._last_epos to SpoolManager. - Update _last_epos on every toolhead status update. - Initialize _last_epos alongside _highest_epos in _handle_klippy_ready (so spool activation cannot rebase to 0 before the first status callback). - In set_active_spool(), after updating self.spool_id, set: self._highest_epos = self._last_epos