drivers: sensor: stm32_vbat: get rid of floating point computation

Instead of using floating point operations to compute the vbat voltage,
it is possible to do the computation using 32-bit variables by
reordering operations and using the sensor_value_from_milli() function.

On a STM32G0, this saves 140 bytes of flash, excluding the FP library
needed on a FPU less MCU.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>