I've implemented vlen-agnostic AES for riscv vector crypto extension (zvkned):
I've chosen to not merge this yet as I don't yet have real hw to measure performance with.