[Megamouse]

• make some parameters const
• cache temporary buffers outside the loop
• use ranged loops
• ensure src buffer size + dst offset don't exceed dst buffer size
• ensure src buffer size + dst offset don't overflow
• use read_from_ptr for key and iv reads to prevent OOB reads when key_idx*16 exceeds the data size
• prevent underflow: use (key_idx < key_count) instead of (key_idx <= key_count - 1)
• fix uncompress:
  • decomp_buf_length can be smaller than the input size after uncompress, meaning that we could end up writing too much data afterwards
  • zlib_buf was a copy of the entire buffer, but then passed zlib_buf.data() + offset and zlib_buf.size() as source and sourceLen, meaning sourceLen was too large.
    Instead, now we copy the segment from offset to end and simply pass zlib_buf.data() and zlib_buf.size() as source and sourceLen