machine/ra4m1: add SCI UART driver - #6015

I tried this PR on an Arduino Uno R4 WiFi (board target arduino-uno-r4-wifi) and the UART works.

Test program:

package main

import "time"

func main() {
	for {
		println("hello")
		time.Sleep(time.Second)
	}
}

Running tinygo flash -monitor -target=arduino-uno-r4-wifi main.go prints hello once a second, with clean output at the default 115200 baud.

The pin and event numbers also match ArduinoCore-renesas 1.6.0 (P109/P110 on SCI9, ELC events 0x9E, 0xA3, 0xA8, PSEL 0x04/0x05).

One issue from reading the code, which I have not tested on hardware. Configure falls back to UART_TX_PIN and UART_RX_PIN for every UART. On the WiFi board those are D22 and D23. UART2.Configure(UARTConfig{}) would give D22/D23 the SCI2 function select, which breaks UART1 and leaves UART2 on the wrong pins. Each UART should have its own default pins.

More testing on an Arduino Uno R4 WiFi (arduino-uno-r4-wifi) with this PR applied. I wrote a set of test programs plus a Python runner that flashes each one, drives the board over the USB serial port with pyserial, and checks the output. The Minima was not tested.

Results:

Issues:

1. 921600 baud does not work. SetBaudRate(921600) computes N=7, which gives about 857 kbaud, 7% below the target. The host receives only 0x80 bytes. The other rates are within +0.2%. Supporting 921600 needs the bit rate modulator (SEMR.BRME and MDDR), or the supported maximum could be documented as 460800.
2. Default pins are not per UART. UART2.Configure(UARTConfig{}) on the WiFi board falls back to UART_TX_PIN and UART_RX_PIN, which are D22 and D23. That gives D22/D23 the SCI2 function select. After that call, UART1 output stops completely, so this is confirmed on hardware. Each UART needs its own default TX and RX pins. UART3 works when its pins are passed explicitly (UART3_TX_PIN and UART3_RX_PIN).

Everything else behaved correctly, including RX interrupts, recovery after an RX buffer overflow, reconfiguring, and operation alongside the scheduler and GC.

I have made some changes covering the issues.
UART now has per-instance default pins which is set in init(), Configure falls back to those instead of UART_TX_PIN and UART_RX_PIN.

I also implemented the bit rate modulator.

UART2.Configure(UARTConfig{}) no longer kills UART1 on the WiFi board, 921600 passes 20/20, and lower rates keep the same N and CKS as before so no change there.

Have not tested Minima, only WiFi board.

Retested the updated commit (9c8f3b04) on an Arduino Uno R4 WiFi (arduino-uno-r4-wifi) with the same test programs and Python runner as before. All tests pass now.

Both issues from my earlier comment are fixed. 921600 now works (N=6, MDDR=236, about 0.03% rate error). UART1 is unaffected by a default UART2.Configure.

One small point. Configure still treats config.TX == 0 && config.RX == 0 as "use the defaults", so an explicit TX of pin 0 can't be selected. That is the same pattern other ports use, so it may be fine.

The Minima is still untested.