process_replay/migration: doesnt need to import acados (#38040)

process_replay/migration: drop transitive dep on long_mpc/acados

migrate_longitudinalPlan only needs get_accel_from_plan + CONTROL_N_T_IDX,
both available from drive_helpers + ModelConstants directly. Importing
from longitudinal_planner instead pulled in long_mpc -> libacados.so as
an eager side-effect, which downstream log-migration consumers (e.g.
commaai/xx pipeline) shouldn't pay for.

selfdrive/test/process_replay/migration.py

@@ -14,7 +14,7 @@ from opendbc.car.gm.values import GMSafetyFlags
 from openpilot.selfdrive.modeld.constants import ModelConstants
 from openpilot.selfdrive.modeld.fill_model_msg import fill_xyz_poly, fill_lane_line_meta
 from openpilot.selfdrive.test.process_replay.vision_meta import meta_from_encode_index
-from openpilot.selfdrive.controls.lib.longitudinal_planner import get_accel_from_plan, CONTROL_N_T_IDX
+from openpilot.selfdrive.controls.lib.drive_helpers import CONTROL_N, get_accel_from_plan
 from openpilot.system.manager.process_config import managed_processes
 from openpilot.tools.lib.logreader import LogIterable
 
@@ -126,7 +126,7 @@ def migrate_longitudinalPlan(msgs):
     if msg.which() != 'longitudinalPlan':
       continue
     new_msg = msg.as_builder()
-    a_target, should_stop = get_accel_from_plan(msg.longitudinalPlan.speeds, msg.longitudinalPlan.accels, CONTROL_N_T_IDX)
+    a_target, should_stop = get_accel_from_plan(msg.longitudinalPlan.speeds, msg.longitudinalPlan.accels, ModelConstants.T_IDXS[:CONTROL_N])
     new_msg.longitudinalPlan.aTarget, new_msg.longitudinalPlan.shouldStop = float(a_target), bool(should_stop)
     ops.append((index, as_reader(new_msg)))
   return ops, [], []