stm32/pyb_can: Fix STM32G4 FDCAN source clock frequency.

Fixes automatic baudrate calculation results.

Default clock source on this SoC is HSE not PCLK1.  We could fix this by
switching to PCLK1 instead, but two extra complications:

- PCLK1 on this board is a 42.5MHz and the Pyboard CAN sample_point
  calculation requires an exact match, which is harder to hit with this
  source frequency.

- Would be a breaking change for any existing Python code on this board,
  i.e. specifying brp, bs1, bs2 to initialise CAN.

In the future it might be worth looking switching to the PLL source on this
SoC instead, as this is a much higher frequency that would give higher
quality BRS bitrate matches (probably too high without using the second
divider going into the CAN peripheral though, so more code changes needed
also).

This work was funded through GitHub Sponsors.

Signed-off-by: Angus Gratton <angus@redyak.com.au>

ports/stm32/pyb_can.c

@@ -182,7 +182,10 @@ static uint32_t pyb_can_get_source_freq() {
             break;
         }
     }
-    #else // F4 and F7 and assume other MCUs too.
+    #elif defined(STM32G4)
+    // STM32G4 CAN clock from reset is HSE, unchanged by MicroPython
+    can_kern_clk = HSE_VALUE;
+    #else // G0, F4, F7 and assume other MCUs too.
     // CAN1/CAN2/CAN3 on APB1 use GetPCLK1Freq, alternatively use the following:
     // can_kern_clk = ((HSE_VALUE / osc_config.PLL.PLLM ) * osc_config.PLL.PLLN) /
     //  (osc_config.PLL.PLLQ * clk_init.AHBCLKDivider * clk_init.APB1CLKDivider);