VoodooI2C matches Intel LPSS I2C controllers by PCI ID, one personality per PCH generation, each carrying the bus clock and the SDA/SCL timing values for that silicon. 2.9.1 knows Sunrise Point, Cannon Lake, Comet Lake and Ice Lake. The Arrow Lake-HX / 800-series PCH functions 8086:7F4C, 7F4D, 7F4E and 7F4F are not in the list, so nothing attaches and a touchpad on those buses never appears.
Linux drivers/mfd/intel-lpss-pci.c assigns these four IDs bxt_i2c_info: clock 133 MHz, SDA hold 42 ns, SDA falling 171 ns, SCL falling 208 ns. VoodooI2C's Ice Lake personality (0x34E8, also bxt_i2c_info in Linux) carries exactly those numbers, so the personality here is the Ice Lake one with the IOPCIMatch changed to 0x7f4c8086&0xFFFCFFFF (mask the two low bits of the device ID, so one entry covers all four functions).
Version 1.0.0 used VoodooI2CPCIController as the class and did not work: the controller enumerated, VoodooI2C attached, and every register read returned 0xffffffff, including COMP_TYPE, which the driver checks first. That is what a PCI function in D3hot looks like: the firmware had left the unused I2C controller powered down, and nothing had brought it back. VoodooI2C's VoodooI2CPCILakeController subclass (used for its Ice Lake and newer personalities) writes the power management control register (PMCSR, in the capability at offset 0x80) to put the function in D0 before it maps the registers. Switching the class to that one is the whole of 1.1.0, and the touchpad has worked since.
The touchpad's interrupt line is a GPIO pin. VoodooGPIO routes those for the PCH generations it has community data for; Arrow Lake's GPIO controllers (INTC1082, INTC1084) are not among them. With -vi2c-force-polling VoodooI2C polls the device instead, which is what runs here. Adding the Arrow Lake GPIO community tables to VoodooGPIO (from the Linux pinctrl-meteorlake / Arrow Lake pinctrl driver) is the proper fix and is not done.
ioreg -r -c VoodooI2CPCILakeController -l -w0
shows the controller with the four timing properties and, below it, the device nub and VoodooI2CHIDDevice. If the controller is there but the register dump shows 0xffffffff, the D0 transition did not happen; if nothing matches at all, check that IOPCIMatch covers your device ID (ioreg -r -n I2C1 -l | grep device-id).