wifimodeswitch: Support also modern A12/k4.19+ MTK HAL

[JamiKettunen]

Android 12+ / kernel v4.19+ modern MediaTek SoCs like MT6768/6789/6877 found on Volla devices do not provide a libwifi-hal.so but instead a libwifi-hal-mtk.so which provides the same functionality while the current fallback libhardware_legacy.so doesn't work.

There still seems to be some weirdness to do with the wlan1 interface magically becoming permanently managed (despite https://gitlab.com/ubports/porting/reference-device-ports/android11/volla-phone-22/volla-mimameid/-/blob/master/overlay/system/lib/udev/rules.d/99-mtk-wlan1-unmanaged.rules) after toggling hotspot even once and rebooting etc. Will see how this behavior seen on 24.04 with this PR fairs to what the status quo is on 20.04 but it's certainly odd..

[JamiKettunen]

Ok, so on UT 24.04 (but not 20.04) if you toggle it even once either as-is currently (with it failing) or with this patch already applied on a clean install you'll get wlan1 managed after reboot going forward despite the udev rule applying fine as per udevadm info /sys/class/net/wlan1 returning E: NM_UNMANAGED=1.. since that's possible to happen in the first place and the following seems to fix it I'll just migrate all affected Volla devices to use it instead and not think about what sort of weird state file mess is going on here:

$ /etc/NetworkManager/conf.d/unmanaged-wlan1.conf
[device-wlan1-unmanaged]
match-device=interface-name:wlan1
managed=0

[peat-psuwit]

(Mirroring my comment from UT packaging MR: https://gitlab.com/ubports/development/core/packaging/mechanicd/-/merge_requests/3#note_2811084177)

I guess we can have this as a distro patch in UT, but not in the upstream repo itself. The reason is that this is SoC-specific private API which coincidentally have the same API as a legacy HAL API. Mechanicd should not to have vendor-specific stuffs in it.

The proper way to do this is probably using HAL over binder instead. This should work independent of vender, but would entail using libgbinder which is somewhat cumbersome I think.

As for UT, since it's a simple, non-invasive fix, we can have this as a backportable hotfix. Thank you for not adding it back to lomiri-indicator-network BTW.

[fredldotme]

Since I created mechanicd, I might as well chime in. IMO it's perfectly fine to merge this change here, it's an isolated addition to what already exists. Now I agree we should probably add Binder support for other devices, but this is okay.