Make ringbuffer more robust or improve usage comments

[axel-h]

Follow-up from #18 (comment)

The ring buffer does not support 0 as data value, but expects this as an end marker. Thus I wonder if rb_new()should do a r->base[0] = 0 to ensure it starts with an empty buffer. Or do a check that the buffer contains at least one NULL value of pre-filled buffers are also supported. Otherwise rb_poll_byte() would loop forever in the worst case.

[kent-mcleod]

Maybe we consider just removing the library before merging the libsharedringbuffer implementation?