i8256: replace diy clock with device_serial_interface and deal with interrupts

[stonedDiscord]

From #14478

Also reenables savestates.

[cuavas]

device_serial_interface is useful for high-level simulation of devices that use asynchronous serial communication. However, you usually end up digging a hole for yourself if you try to implement detailed UART behaviour on top of it (or God forbid USART behaviour). As such, I'd consider making a UART device more dependent on device_serial_interface to be regressive.

[stonedDiscord]

I copied the old implementation from
https://github.com/mamedev/mame/blob/master/src/devices/machine/i8251.cpp#L139
under the assumption that Intel didn't reinvent the wheel.
During testing in the last months it turned out it wasn't working and the one in this PR is working.
The datasheet isn't very thorough on the internals and I don't want to boil my working 8256 in acid so I'm unsure how to do this accurately.

[cuavas]

MAME’s 8251 is a terrible starting point, as it’s in dreadful shape. You also copy/pasted a pile of synchronous mode support code which wasn’t necessary But if you look at the history of that file, you’ll see that the device_serial_interface dependency has been reduced over time, not increased. A simple UART is pretty easy to implement directly anyway. It’s mostly a couple of shift registers and counters. The complicated part is the host interface, which device_serial_interface doesn’t help you with.

Anyway, no increased dependency on device_serial_interface in UART devices.

[ajrhacker]

No, the 8256's asynchronous communications interface doesn't work very much like the 8251 USART.