rp2: use the correct channel mask for rp2350 ADC; hold lock during read

[eliasnaur]

Found while investigating https://github.com/orgs/tinygo-org/discussions/4936.

Untested.

[soypat]

Why are we locking? Is this due to new multicore semantics?

[eliasnaur]

I guess so. And because the ADC is shared between the analog pins, it's reasonable to support separate goroutines using it simultaneously.

[soypat]

Either we test it or we could do a side-by-side comparison of existing pico-sdk code... or a datasheet walkthrough would work

[eliasnaur]

It's tested here: https://github.com/orgs/tinygo-org/discussions/4936#discussioncomment-13603066.

[soypat]

I've noted it is tested but exhibits no change to behaviour so I'm still curious as to what this CL solves.

[eliasnaur]

The patch does 2 things:

• Move the reading of the ADC inside the lock, on the basis that if the lock is needed, the actual reading the value should be locked as well.
• Use rp.ADC_CS_AINSEL_Msk instead of the hard-coded 0b111 mask, which is wrong for rp2350.

[soypat]

Alright, LGTM in that case! Thank you for the contribution @eliasnaur !