RMT: Various cleanups, small API enhancements

[wisp3rwind]

Thank you for your contribution!

We appreciate the time and effort you've put into this pull request.
To help us review it efficiently, please ensure you've gone through the following checklist:

Submission Checklist 📝

• I have updated existing examples or added new ones (if applicable).
• I have used cargo xtask fmt-packages command to ensure that all changed code is formatted correctly.
• My changes were added to the CHANGELOG.md in the proper section.
• I have added necessary changes to user code to the latest Migration Guide.
• My changes are in accordance to the esp-rs developer guidelines

Extra:

• I have read the CONTRIBUTING.md guide and followed its instructions.

Pull Request Details 📖

Assorted small-ish refactorings (sorry for the large number of commits, but they should all be fairly self-contained mechanical changes):

• most commits are purely internal, to make things more consistent, easier to reason about, and remove slight inefficiencies

• ensure that all methods that take channels by value also return them under all circumstances (instead of consuming them permanently on error, which some of them did)

• rename SingleShotTxTransaction to TxTransaction since the previous name was quite unwieldy, and this also better shows the symmetry between RxTransaction and TxTransaction

Testing

HIL tests.

[wisp3rwind]

Actually, hold on - the enumset usage in 964418a seems to blow up some closure sizes, which in turn won't be inlined, leading to bad codegen...

[wisp3rwind]

Should be good now. What do you think?

---

Actually, hold on - the enumset usage in 964418a seems to blow up some closure sizes, which in turn won't be inlined, leading to bad codegen...

What happened is that I had rewritten

#[inline]
pub fn clear_tx_interrupts(&self) {
    let rmt = crate::peripherals::RMT::regs();

    rmt.int_clr().write(|w| {
        w.ch_tx_end(self.ch_idx as u8).set_bit();
        w.ch_tx_err(self.ch_idx as u8).set_bit();
        w.ch_tx_loop(self.ch_idx as u8).set_bit();
        w.ch_tx_thr_event(self.ch_idx as u8).set_bit()
    });
}

to

#[inline(always)]
pub fn clear_tx_interrupts(&self, events: impl Into<EnumSet<Event>>) {
    let rmt = crate::peripherals::RMT::regs();
    let events = events.into();
    let ch_idx = self.ch_idx as u8;

    rmt.int_clr().write(|w| {
        if events.contains(Event::End) {
            w.ch_tx_end(ch_idx).set_bit();
        }
        if events.contains(Event::Error) {
            w.ch_tx_err(ch_idx).set_bit();
        }
        if events.contains(Event::LoopCount) {
            w.ch_tx_loop(ch_idx).set_bit();
        }
        if events.contains(Event::Threshold) {
            w.ch_tx_thr_event(ch_idx).set_bit();
        }
    });
}

which

1. obviously increased the amount of code in the closure but
2. increased it even more then intended: self.ch_idx is a ChannelIndex enum with values such that bounds checks from the PAC are eliminated. However, moving the cast out of the closure loses that information when compiling the PAC accessors.

In the end, the compiler chose not to inline the closure anymore, such that the bounds checks (one for each bit!) remained in the binary and the code is not optimized as part of the caller where the events argument is always known at compile time.

I've now moved the cast to inside the closure, and also added #[inline] attributes on the closure, which should avoid such issues.

[bugadani]

Could you please move the changelog/migration guide to the next version?