Allow encode_into arbitrary buffers. - #1160

Lets Encoder.encode_into() (both msgspec.msgpack and msgspec.json) accept any writable, contiguous buffer-protocol object (e.g. a memoryview onto multiprocessing.shared_memory.SharedMemory), not just bytearray.

Currently the encode_into only works on bytearrays, this makes it impossible to use Python's SharedMemory with a zero-copy serialization.

from multiprocessing import shared_memory
import msgspec

enc = msgspec.msgpack.Encoder()
shm = shared_memory.SharedMemory(
    name="shared_data",
    create=True,
    size=1024,  
)

enc.encode_into(Message(datetime.now().astimezone(), Severity.ERROR, "TEST"), shm.buf, 0)

fails with

TypeError: buffer must be a `bytearray`

This changes allows for arbitrary buffers that do not support resizing.

encode_into also return the number of newly written bytes. So offset + newly_written_bytes can by used as the offset for the next element when streaming data.