Allow USB to work with V2 clocking API

[rnd-ash]

Summary

Continuing my work to allow more peripherals to work with the V2 clocking API, this PR adds a new method (Maybe name should be changed to something more permanent), that allows the UsbBus to be constructed with the new V2 clocking API.

I've left the original constructor untouched in order to maintain compatibility.

Now, in the BSPs, USB can be constructed like so:

pub fn usb(
    usb: pac::Usb,
    pclk_usb: Pclk<Usb, Gclk0Id>, // Can use any Gclk here
    mclk: &mut pac::Mclk,
    d_minus: impl Into<UsbDm>,
    d_plus: impl Into<UsbDp>,
) -> UsbBusAllocator<UsbBus> {
    UsbBusAllocator::new(UsbBus::new_with_v2_clock(pclk_usb, mclk, d_minus.into(), d_plus.into(), usb))
}

[rnd-ash]

Additional note. I'm not sure why we even need the pclk here, since the original constructor never made use of the UsbClock that it was given. Maybe we can remove the clock entirely from the V1 constructor and then it would work with both clocking APIs?

[jbeaurivage]

@rnd-ash, now that I have a better grasp of the v2 clocking system I feel more comfortable commenting on this:
The design philosophy behind clocks::v2 is that the entire clock tree should be configured by the user. This includes the main clock, but also enabling the PCLKs and AHB/APB clocks as well.
Case in point, the MCLK peripheral isn't even given out when creating the tokens with clocks_at_reset. I think one would need to use unsafe code just to use new_with_v2_clock, breaking the guarantees laid out by clock::v2.

The types being passed is just a way of proving that the clocks are in fact correctly enabled, at compile time. What's more, these types must be kept around in the USB struct until it's dismantled (ie, give ownership to the peripheral struct), because otherwise one could "prove" the clocks are setup correctly when setting up a peripheral, then immediately go and use those types to disable the clocks, rendering the peripheral unusable.

[rnd-ash]

When I created this PR, I was learning the crate, and I didn't realise until recently that you can call pclk.into() to create a USB clock that the V1 API requires.

Therefore, I believe it's safe to close this PR as it is semi pointless