i2c: implement I2C:transaction for embedded-hal and embedded-hal-async

[liebman]

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).
• My changes were added to the CHANGELOG.md in the proper section.

Extra:

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

Pull Request Details 📖

Description

This is an implementation of the I2C::transaction method for embedded-hal and embedded-hal-async. Also added native async read/write/write_read methods not requiring embedded-hal/embedded-hal-async

This fixes #1490

Testing

I've tested:

• esp32 - sync and async
• esp32s2 - sync and async
• esp32s3 - sync and async
• esp32c2 - sync and async
• esp32c3 - sync and async
• esp32c6 - sync and async
• esp32h2 - not tested (I don't have one)

sync testing was i2c_display and i2c_bmp180_calibration_data examples
async testing was embassy_i2c_bmp180_calibration_data example (new)

[bjoernQ]

Awesome - this works now as expected in regards to #1490

Push the "transaction" down into the lower level implementations so that the read/write/write_read (non-embedded-hal trait version) can use the same code rather than a similar but different implementation.

I think that would be the best way to go. We also want to make every driver useable without depending on embedded-hal following the Focus on Drivers idea in https://blog.rust-embedded.org/embedded-hal-v1/

I remember we had some issues with the C2 before because the FIFO is only 16x8 bits there (while IIRC all others are 32x8) - we especially had problems with SSD1306 displays - one PR for that was #213 (I think there were follow up PRs)

In general, this will be a great improvement - Thanks for taking care of this!

[liebman]

Learned something new for the esp32: The End command never gets its cmd_done bit set! The end_detect interrupt bit gets set and even after a 1ms delay, the cmd_done bit for that is not set. (This is not mentioned in the TRM or the errata).

[bjoernQ]

Amazing! Works great for me

I think I remember we usually don't limit the blocking impls to the blocking mode. Did that cause problems here?

[liebman]

Yes - It prevented me to from adding read/write/write_read for Async (not using embedded-hal-async)

[bjoernQ]

LGTM - Thanks!

[JurajSadel]

LGTM, thanks!