improve GATT for bluetooth, improve performance, stability for ble transport

[0xkelvin]

should be: ~30KB/s - for #3432
Gatt performance improved, so it should be solved this issue as well : #2782

Please help to test again in your side

[gemini-code-assist]

Code Review

This pull request refactors the GATT transport layer to improve performance and stability by aggregating smaller packets into larger ones before sending. While the approach is sound, I've identified three critical bugs in the implementation. These include an incorrect buffer space check that could lead to overflows, faulty error handling when reading from the queue, and an incorrect payload length calculation that would cause data truncation. These issues must be fixed to ensure the correctness and stability of the BLE transport.

[beastoin]

@0xkelvin man, thank you. where are you based? I'd like to send you and omi + fpc cable so you can freely dev/test, no blocker.

[0xkelvin]

@beastoin I am base in Singapore, let me DM you on Discord for my address

[0xkelvin]

@beastoin

after improvement on devkit , ble GATT reaches around 182 kbps =~ 23KB/s

before improvement on devkit, ble GATT reaches around 38.2 kbps =~ 5KB/s

[beastoin]

i see, will send you one @0xkelvin

the speed is good. btw:

1/ does your changes change the structure of the audio packet? that requires the mobile app's updates. pls note the structure of the packet before | after.

actually the #3432 relates to the storage syncs (check storage.c) but not real time audio sync (pusher)

[0xkelvin]

@beastoin
Still room to improve transport by sending multi opus encoded frame on 1 mtu package, which require the phone app update according to new packet structure. .. Working on it

[beastoin]

take your time, masked as draft fn, lmk once it is ready