[export_arm_cortexm] [fix] Standardize 'sub' operator naming to 'aidge_sub_float32' in export_arm_cortexm

Within export_arm_cortexm, the 'sub' operator was referred to using different names depending on the context. To ensure the export compiles correctly and to match the naming convention used for 'add' (i.e., aidge_add_float32), I standardize all calls to use 'aidge_sub_float32'.