mm_port_subsys_to_qmi_endpoint_type() should not assume that WWAN == PCIe

Moved here from !549 (merged); it doesn't seem to cause any problems at the moment so we don't necessarily need to solve this before merge.

mm_port_subsys_to_qmi_endpoint_type() contains this comment at the moment:

    /* ... */
        case MM_PORT_SUBSYS_RPMSG:
        case MM_PORT_SUBSYS_QRTR:
            return QMI_DATA_ENDPOINT_TYPE_EMBEDDED;
        /* The WWAN subsystem abstracts the underlying transport bus, and so
         * endpoint type can not be deducted from that. This function should
         * then be revisited, but in practice, only MHI/PCI modem ports are
         * exposed through the WWAN subsystem for now.
         */
        case MM_PORT_SUBSYS_WWAN:
            return QMI_DATA_ENDPOINT_TYPE_PCIE;
    /* ... */

This is no longer true. With https://lore.kernel.org/netdev/20210618075243.42046-1-stephan@gerhold.net/ RPMSG can also go through the WWAN subsystem now. In that case we would want to return QMI_DATA_ENDPOINT_TYPE_EMBEDDED. We need some way to better detect the kind of device that provides the WWAN ports.

For now this does not seem to cause any problems, I suspect the endpoint type does not actually end up being used for RPMSG+BAM-DMUX at the moment. ctx->use_endpoint in mm-port-qmi.c shouldn't end up being set for the RPMSG+BAM-DMUX case but if this changes at some point we'll need to fix this.