Ambiq Apollo3: Make SFE_ARTEMIS use arduino uno form factor, fix GPIO IRQ API

[multiplemonomials]

Summary of changes

I'm starting to test Ambiq Apollo3 on the CI test shield. In order to do this, I first had to add arduino uno form factor support for the SFE_ARTEMIS board. This board does use the physical arduino uno form factor, but was not configured as such in Mbed, meaning the test shield project wouldn't compile. I updated PinNames.h to fix this.

With that sorted out, I investigated some failures I was seeing in the InterruptIn test. Turns out that there were two bugs in this target's implementation:

• For some god-awful reason (I think due to reuse of register fields for multiple meanings), the Apollo3 does not allow you to read a GPIO pin normally if it's configured to have interrupts disabled. This meant that calling InterruptIn::read() after constructing the object but before calling rise() or fall() would always return 0 instead of the actual logic level on the pin
  • Yes, this means that in normal operation interrupts are actually enabled but masked on every single GPIO pin
• Incorrect logic in the HAL meant that if you enabled both rise and fall interrupts on a pin, it would call InterruptIn::_irq_handler() with IRQ_RISE | IRQ_FALL as the event argument, which is not valid and led to no callback getting executed.

To fix the first issue, I set up a workaround where the pin is kept in interrupts-enabled-but-masked mode when no interrupt is configured, so that it can be read. To fix the second issue, I updated the HAL to read the pin state in the ISR and then send the rise or fall event accordingly. This is a little bit janky, but should work OK in most cases, and it's how STM32 HAL currently works.

I also realized that there was a big issue with how this family of boards writes its PinNames.h files: it only contains "Dx" and "Ax" names, which match up with the labeling on the board, but do not map consistently to MCU pins. This makes it a pain to write pinmaps for this target, because you constantly have to cast integers into PinNames if you want to reference the MCU pin numbering.

I updated all of the PinNames.h files to use "IO_xx" to mean pin XX of the MCU, and then map the hardware names to those IO_xx values instead of just integers. This should not be a significant compatibility change (apart from names getting standardized like SW1 -> BUTTON1), but does make code for this MCU a lot clearer.

Impact of changes

• InterruptIn::read() now works on Apollo3 if no interrupt is configured
• You can now set rising and falling edge callbacks on a single pin on Apollo3 and they will both get called
• All Apollo3 PinNames.h files updated to Mbed 6 standards (meaning Mbed knows about the LEDs on the boards now)

Migration actions required

Documentation

None

---

Pull request type

[X] Patch update (Bug fix / Target update / Docs update / Test update / Refactor)
[] Feature update (New feature / Functionality change / New API)
[] Major update (Breaking change E.g. Return code change / API behaviour change)

---

Test results

[] No Tests required for this change (E.g docs only update)
[] Covered by existing mbed-os tests (Greentea or Unittest)
[X] Tests / results supplied as part of this PR

InterruptIn test shield test now passes (except some weird analog propagation time related stuff that I still need to debug).

---