extmod/asyncio: Add a native asyncio REPL

[andrewleech]

Summary

This adds asyncio.arepl, an interactive REPL that runs as an asyncio task instead of blocking the event loop, so other tasks (Wi-Fi, USB host/device, a socket server, etc) keep running while it waits for and executes input.

Rather than reimplementing line editing in Python, it drives the existing event-driven C REPL for paste, raw, raw-paste and continuation handling, and only defers a completed line to the event loop when it contains a top-level await. That relies on the compiler's existing MICROPY_COMP_ALLOW_TOP_LEVEL_AWAIT support (already used by the webassembly port), which wraps such a line in a generator so calling it returns a coroutine - no VM changes needed to make a REPL line awaitable. The blocking raw REPL (used after a soft reset) also now feeds through this same event-driven core instead of carrying a second copy of the raw-REPL protocol.

task() takes stop_loop_on_exit and persistent options, so it can be the sole boot REPL (Ctrl-D soft-resets) or one task among several that stays up for the life of the program. Ctrl-C at the prompt cancels an in-flight await rather than just interrupting the next line. A separate asyncio.arepl.breakpoint() drops into a blocking nested REPL from async code for debugging; its C support is behind its own MICROPY_REPL_ASYNCIO_BREAKPOINT flag (off by default below the EVERYTHING ROM level) since it's a debugging convenience rather than part of the core REPL.

It's on by default at the EXTRA_FEATURES ROM level. A few boards explicitly disable it rather than drop ROM level: PYBD_SF2/SF3 and the esp8266 1M/512K variants are too flash-tight, and zephyr/qemu's stdin isn't pollable. PYBD_SF6 picks it up newly since it previously inherited SF2's disable through shared base config despite having 2 MiB of flash to spare.

Testing

Tested on unix: the cmdline REPL suite (including a new repl_asyncio_intr.py covering Ctrl-C cancellation of an in-flight await) passes, plus the wider test suite and pre-commit's lint/format/spelling checks. Build-verified for stm32 PYBV10 and PYBD_SF6 (flash-size impact below); rp2 RPI_PICO didn't build in this environment due to an unrelated local picotool/CMake version-detection problem, so it's CI-verified there rather than locally rebuilt this round. Earlier hardware testing on an RP2350 (REPL as one task alongside a Wi-Fi supervisor and socket-accept loop, UART and socket REPL both surviving socket churn) predates the later rework that switched arepl.py to drive the existing C REPL instead of reimplementing line editing in Python, so that run validates the architecture rather than this exact revision.

Trade-offs and Alternatives

Flash cost is about 1.6-1.9 KB on stm32: PYBV10 goes from 374,308 to 376,212 B text+data (+0.51%), PYBD_SF6 from 1,094,952 to 1,096,552 B (+0.15%). Most of that is the extra code paths added to the shared event-driven C REPL plus the new module glue; the frozen arepl.py itself is small (under 1 KB) since it only drives the existing REPL and dispatches top-level-await lines to the event loop, rather than reimplementing paste/raw-mode/continuation handling in Python. That reuse is the main design trade-off here: it costs a bit more flash than a from-scratch Python REPL would, but avoids a second implementation of line-editing/paste/raw-mode that would need to stay in sync with the C one. The breakpoint() debugger REPL is kept out of that default cost by its own flag.

Generative AI

I used generative AI tools when creating this PR, but a human has checked the code and is responsible for the code and the description above.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.54%. Comparing base (13303f8) to head (eba1b1f).
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@                     Coverage Diff                      @@
##           review/native_async_repl      #49      +/-   ##
============================================================
+ Coverage                     98.50%   98.54%   +0.03%     
============================================================
  Files                           177      182       +5     
  Lines                         22927    23332     +405     
  Branches                          0        5       +5     
============================================================
+ Hits                          22584    22992     +408     
+ Misses                          343      339       -4     
- Partials                          0        1       +1     

Flag	Coverage Δ
unix-coverage-32bit	98.54% <100.00%> (?)
unix-coverage-64bit	98.47% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

Code size report:

Reference:  py/py.mk: Move MICROPY_FORCE_32BIT to the Windows port's Makefile. [13303f8]
Comparison: extmod/asyncio: Gate breakpoint() debugger behind its own flag. [merge of eba1b1f]
  mpy-cross:  +560 +0.147% 
   bare-arm:   +28 +0.050% 
minimal x86:   +69 +0.037% 
   unix x64: +6152 +0.713% standard[incl +480(data) +72(bss)]
      stm32: +2940 +0.730% PYBV10[incl +16(bss)]
      esp32: +12848 +0.728% ESP32_GENERIC[incl +2368(data) +48(bss)]
     mimxrt: +2248 +0.571% TEENSY40
        rp2: +12420 +1.307% RPI_PICO_W[incl -540(bss)]
       samd: +2308 +0.831% ADAFRUIT_ITSYBITSY_M4_EXPRESS[incl +44(bss)]
  qemu rv32:  +763 +0.167% VIRT_RV32