#15836: Update reads, writes, and synchronize ttnn apis to take in sub device ids

[tt-aho]

Ticket

#15836

Problem description

Need to expose the new sub_device_ids args for various metal apis up the stack to enable usage in python. The sub_device_ids parameters are mostly used so that user can specify what dispatch should wait/stall on. Ex. If we have a persistent program on one sub-device, and regular programs on another, we'd only want to block/stall on the sub-device running regular programs during runtime, and only block on all programs at the end when the persistent program finishes as well, otherwise if we always stall on all sub-devices we'd hang waiting for the persistent program to finish when we only need the regular program to finish.

What's changed

Update ttnn apis up the stack to propagate sub_device_ids down to various metal apis.

Checklist

• Post commit CI passes https://github.com/tenstorrent/tt-metal/actions/runs/12251118948
• Blackhole Post commit (if applicable)
• Model regression CI testing passes (if applicable) https://github.com/tenstorrent/tt-metal/actions/runs/12251124053
• Device performance regression CI testing passes (if applicable) https://github.com/tenstorrent/tt-metal/actions/runs/12251122016
• New/Existing tests provide coverage for changes

[omilyutin-tt]

Do we have an issue # for this? Also, is there any documentation related to sub devices, or design docs? It isn't very clear how / when to use sub devices.

[omilyutin-tt]

Thanks!

[ayerofieiev-tt]

I hope the team will find time to review these APIs and sort things out.

[blozano-tt]

@tt-aho could this PR be at fault for breaking T3K unit tests?

https://github.com/tenstorrent/tt-metal/actions/runs/12306978859/job/34350073706

cc: @ttmchiou

Running this job at the commit just before your commits, to confirm:

https://github.com/tenstorrent/tt-metal/actions/runs/12309298990