adc improvements for dma

[gernoteger]

This PR should fix adc with dma issues as mentioned in #282: "DMA Transfer consumes peripheral".

Changes:

• DMA Transfer exposes a reference to the peripheral, currently only for PeripheralToMemory
• the ADC provides a closure similar to sample_to_millivolts
• a demonstration of its use in the adc_dma_rtic example.

As of now this PR is not complete; I kindly ask to review the concept before I complete the implementation, documentation and squashing.

[burrbull]

It is interesting way.
Is there sense to convert to u16 implicitly if you already have u32?

Squash, please.

[burrbull]

cc @gernoteger
Rebase and squash

[gernoteger]

It is interesting way. Is there sense to convert to u16 implicitly if you already have u32?
The u16 was also used original sample_to_millivolts, I just didn't want to change maths in this PR. Thinking of it, this conversion (also in the original) may lead to a loss of resolution: when using 12 bit ADCs, we get 4096 possible steps from the ADC, while a 3.3V VRef will, only produce 3300 Steps.

[burrbull]

You've broken git

[gernoteger]

Sorry. I'm confused; should I quit working on this PR, and continue with #406 instead?

[burrbull]

This PR is preferable but it should be 1 - 2 commits rebased on master branch with no Merge.

[gernoteger]

I fixed the git issue. Still, I'd like to tackle the issue with the lost resolution by converting samples to milliseconds. Probably this con go to a different PR, what do you think?
The Changelog still needs some fixing

[burrbull]

DMA code for this HAL is not mine, so I need to clarify is such access safe?

[burrbull]

cc @thalesfragoso

[burrbull]

@gernoteger I still can't merge this as this method gives unsafe access any DMA payload.
Possible solution:
b275233

[gernoteger]

thanks for doing all the work. I incorporated your suggestion nearly 1:1, only fixed a small typo in a comment

[burrbull]

I fixed the git issue. Still, I'd like to tackle the issue with the lost resolution by converting samples to milliseconds. Probably this con go to a different PR, what do you think?

I'm afraid I don't understand what lost resolution you mean. milliseconds? time?

[gernoteger]

When converting a sample to millivolts, with a 3.3V (or less) VRef, we have <=3300 possible steps, while a 12Bit ADC has 4096. Probably adding a function "sample_to_microvolts" and/or direct access to the calibration data would help here. It's related to this PR, but somewhat orthogonal.

[gernoteger]

I documented the resolution issues mentioned earlier in #412. I think we should handle this in a different PR.
From my point of view this PR is feature complete and can be merged.

[burrbull]

bors r+

[bors]

Build succeeded:

• ci (stm32f401, stable)
• ci (stm32f405, stable)
• ci (stm32f407, stable)
• ci (stm32f410, stable)
• ci (stm32f411, stable)
• ci (stm32f412, stable)
• ci (stm32f413, stable)
• ci (stm32f415, stable)
• ci (stm32f417, stable)
• ci (stm32f423, stable)
• ci (stm32f427, stable)
• ci (stm32f429, stable)
• ci (stm32f437, stable)
• ci (stm32f439, stable)
• ci (stm32f446, stable)
• ci (stm32f469, stable)
• ci (stm32f479, stable)
• Rustfmt