In my limited tests this improves decoding speed by around 3%.

It turns out that “seeking to a first sample in a not–first block” is actually very common, because Tangara only seeks to exact seconds and the reference encoder tends to size blocks in a way that makes the first sample in a block likely be the sample that the firmware wants to seek to.

The decoder does use `memset` on the buffer, but according to Espressif that should be fine.

See-Also: https://docs.espressif.com/projects/esp-idf/en/stable/esp32/api-reference/system/mem_alloc.html#bit-accessible-memory
See-Also: https://esp32.com/viewtopic.php?t=1730#p8056