I haven't looked in a little while, but the QIIME 2 DADA2 wrapper script does generally follow the per-sample denoising strategy in your link. The reason it's so terribly slow is that it's using a "universal" build which can't make as much use of your specific CPU architecture as a compiled version can.
As far as importing what you have, you won't be able to use .rds files as those aren't really a file format so much as memory dump.
But you should be able to load them and then write a fasta file and roughly compatible OTU table.