message consolidation and more use of validators

[dhalbert]

Closes #6392.

To reduce firmware size, this PR uses argument validators where possible and consolidates and simplifies various error messages.

• I went through all the mp_raise calls, and used validators from argcheck.c to simplify the validation. A few new validators were added, building on keypad: support for vector and matrix key scanning #4891.
• Added mp_arg_error_invalid(qstr), which raises ValueError with the message "Invalid %q". This is useful many places, where there were previously more specialized but not necessarily more informative messages. Then also added raise_ValueError_invalid_pin() and raise_ValueError_invalid_pins(), which refactor mp_arg_error_invalid(MP_QSTR_pin) and mp_arg_error_invalid(MP_QSTR_pins), both of which are very common (a couple dozen uses each).
• extmod/vfs_fat.c: refactor multiple identical calls to mp_raise_OSError()
• ps2io: regularize using clock* instead of clk*, to make argument names and internal names more consistent.
• Shorten some mp_raise_NotImplementedError()
• Used %q args to refactor and share some messages.
• Use m_malloc_fail() instead of more chatty specialized messages which were not necessarily more helpful. Thanks @tannewt for pointing this routine out.
• PixelBuf.c: refactor some identical error messages to reduce code size.

I did not try to squeeze the last drop out of various _bleio messages, since we are not short on code space in those builds, and the savings would be minimal.

Some initially measured byte savings (current savings are a few bytes more than this due to further changes):

Board	Savings in bytes
Trinket M0	316
Metro M0	424
CPX	420
Metro M4	1420
Feather nRF52840	924
Feather RP2040	4524
Feather STM32F405	3948
Metro ESP32-S2	4176
Feather i.MX 1062	5632

I think some of the larger differences here are because some builds are not using LTO.

The number of distinct messages in circuitpython.pot went from 1061 to 989.

[gamblor21]

Great work! I had one comment where I think an if may still be able to be removed but I may be wrong. I didn't see anywhere else where the values changed.

I wonder if there may be a few new questions where it does not say why a pin is invalid (e.g. not a DAC pin) and now just says invalid. But probably worth the space savings.

[dhalbert]

I wonder if there may be a few new questions where it does not say why a pin is invalid (e.g. not a DAC pin) and now just says invalid. But probably worth the space savings.

If you look at the backtrace and the line in question, I think it will be clear it's invalid because of the context. There were already cases where the messages just said "Invalid pin" without further explanation; this just takes advantage of regularizing that.

[jepler]

Thanks! I think it could go in just like this, but there are a few items I spotted along the way.

[dhalbert]

With the latest round of changes, Trinket M0 is now up to 544 bytes saved.

[tannewt]

Awesome work! Let's merge this and tag 8.0.0-alpha.0 since it changes some exception types.

[dhalbert]

Let's merge this and tag 8.0.0-alpha.0 since it changes some exception types.

I agree. Then we can merge the postponed 8.0.0 PR's, and implement the hanging 8.0.0 issues.