Spi master: remove public hidden APIs, add Config/apply_config

[bugadani]

Not the nicest PR, I didn't take the concept to its extremes yet. But the doc(hidden) APIs are gone so that might be a valid first step.

cc #2416 and #1919

There are a few bugfixes included as well.

[bugadani]

What should we do with interrupts? We only expose GDMA DMA interrupt flags, not SPI peripheral interrupt flags (meaning none for PDMA devices). IMHO since set_interrupt_handler refers to peripheral interrupts, we should expose the peripheral interrupt flags. However, the underlying DMA's interrupt flags and handler are inaccessible.

[Dominaezzz]

However, the underlying DMA's interrupt flags and handler are inaccessible.

This is part of why I want to split them out into a separate object.

[bugadani]

Yeah a separate .peripheral_interrupts()/.dma_interrupts() returning &InterruptControl would work. But returning either of them isn't trivial - which one to return, what if the user wants both? Do we end up with a single split_interrupt_control that returns all, or multiple ones? Neither of them sound ideal.

[Dominaezzz]

Do we end up with a single split_interrupt_control that returns all

This is probably best imo. At least for the PDMA case where the interrupt registers are shared between the DMA and peripheral. I definitely don't see this being a chip-agnostic API.

[bjoernQ]

LGTM

[bugadani]

mmm I removed uart::config to allow uart::Config, but spi::master::Config/i2c::master::Config is somewhat annoying 😅 oh well...

[MabezDev]

mmm I removed uart::config to allow uart::Config, but spi::master::Config/i2c::master::Config is somewhat annoying 😅 oh well...

at least master/slave adds some context, the config prefix on UART was just extra characters to type :D.

[MabezDev]

LGTM, lets address my other review comments seperately.

[bugadani]

I have elected to pick a random option and resolve the CI warnings. We'll have to touch the code anyway if we figure out what to do about #2416 so I'll take the tech debt and move on.