Add selective collect to memory allocations

[tannewt]

By selectively collecting an allocation, we can skip scanning many allocations for pointers because we know up front they won't have them. This helps a ton when large buffers are being used and memory is slow (PSRAM). In one Fruit Jam example GC times drop from 80+ms to ~25ms. The example uses a number of bitmaps that are now no longer scanned.

[tannewt]

I have this branch with debugging prints and checks: https://github.com/tannewt/circuitpython/tree/debug_gc_selective_collect

[tannewt]

Hopefully #10269 will make enough space for this.

[dhalbert]

This is a big change but is a very good idea.

I tried to do complete // CIRCUITPY-CHANGE labelling, partly for my own benefit at the next upstream mege, and partly for the reader.

As I discussed in two comments, I think it might be safer to force conscious choice between the _with_collect() and _without_collect() versions of things by eliminating the old ambiguous name, or else make the old name do the safer thing of assuming collecting.

[tannewt]

I've also modified MICROPY_ALLOC_GC_STACK_SIZE from 64 to 128. It alone drops GC times from 200+ms to ~110ms without selective collect. Selective collect without the stack increase drops to ~25ms. With the stack increase it drops further to ~15ms.

This is when testing https://github.com/adafruit/Adafruit_Learning_System_Guides/pull/3013/files on Fruit Jam (RP2350 with PSRAM).

[dhalbert]

Looks good! And the m_malloc_without_collect() is clearer.

[AimSD23nSiR]

Jammed

[tannewt]

I wrote up more info about this here: https://blog.adafruit.com/2025/04/28/improving-garbage-collection-time-in-circuitpython/