Micropython v1.27 merge

[dhalbert]

Merge MicroPython v1.27. This merge includes v1.26 as well.

• Upstream extmod/modre.c includes support for pos and endpos arguments in match() and search(). This replaces our own implementation with upstream.
• time epoch handling is now more like our own, also removing some differences.
• MP_NORETURN is now used everywhere instead of NORETURN.
• I fixed things related to ports/unix/coveragecpp.cpp so it is more like upstream.
• ports/unix/main.c is now more like upstream.
• Our own previous MICROPY_NONSTANDARD_TYPECODES now exists in upstream as MICROPY_PY_STRUCT_UNSAFE_TYPECODES.
• MICROPY_PY___FILE__ is now MICROPY_MODULE___FILE__.
• Floating point conversion to/from strings was improved in upstream.
• Support for old Python versions was removed from upstream scripts. We had already done this.
• Some iterator handling we had for arrays is now included in upstream.
• objlist.c list_subscr() is now more like upstream.
• shared/runtime/pyexec.c had a number of changes in upstream. We had diverged noticeably from it already. I spent time getting back closer to the upstream version.
• I added \$ to some trailing whitespace in test .py.exp files so the whitespace would not be trimmed away, and noted this upstream: Handle trailing whitespace in test .exp files using regexp syntax micropython/micropython#19068 and tests/**/*.py.exp: protect trailing spaces with \$ micropython/micropython#19073.

In general, this was a pretty easy merge. It makes me want to submit a few more PR's upstream to remove other differences.

Before submitting this PR, I did a few successful builds and did some REPL smoke tests. I will start this as a draft PR and do some more in-depth testing on boards after all the builds are successful.

[dhalbert]

@tannewt could you take a look at the zephyr test failures? I am not sure what is going wrong here. Certainly there is something but maybe you recognize the issue right away.

I merge from main last night so this includes the latest zephyr code.

I wonder if the zephyr tests should be run on a separate track by the scheduler. This would allow the non-zephyr builds to run to completion instead of being blocked by failing zephyr tests.

[tannewt]

One follow up question that we don't need to answer now.

[dhalbert]

I tested as follows:

• espressif MagTag: tested deep sleep PinAlarm, and separately, a portal weather application.
• atmel-samd PyPortal - tested a PyPortal application that uses the Internet. Also did blink on SAMD21.
• nordic Feather nRF52840 - tested BLE UART
• raspberrypi Fruit Jam: tested Fruit Jam OS. There was some flakiness with the SD card at one point, but it went anyway. It is not clear me this is a new problem. It may have to do with the USB SD card presentation.
• stm32 STM32F405 CIRCUITPY does not work, but neither does it work on 10.1.4. See STM32F405: no CIRCUITPY #10940. So not a new bug.